Add files using upload-large-folder tool
Browse files- README.md +32 -0
- SHA256SUMS +5 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/aistpp_edge_metrics.json +44 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/artifact_samples.jsonl +512 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/audio_metrics.json +26 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/benchmark.json +1058 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/dataset_metrics.json +382 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/merge_summary.json +17 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/metrics.json +1061 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/physical_metrics.json +24 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/sample_metrics.jsonl +0 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/samples.jsonl +0 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_001/samples.jsonl +64 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_002/samples.jsonl +64 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_004/samples.jsonl +64 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_005/samples.jsonl +64 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_006/samples.jsonl +64 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_007/samples.jsonl +0 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shards.json +72 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/summary.md +10 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed.json +565 -0
- benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/transition_metrics.json +5 -0
- benchmark/samples/mixed_modalities_all_v1/.audio_120f_export/audio/manifest.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/.audio_120f_export/export_summary.json +0 -0
- benchmark/samples/mixed_modalities_all_v1/audio_test_512.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/audio_test_512_120f.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/gmr_smplx/smplx/manifest.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/gmr_smplx_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/keypoints_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/materialization_summary.json +14 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/nmr_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/omniretarget_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs/phc_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_phc_zup_fixed/phc_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/keypoints_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/materialization_summary.json +14 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/nmr_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/gmr_smplx/smplx/manifest.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/gmr_smplx_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/keypoints_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/materialization_summary.json +13 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/nmr_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/omniretarget_input_samples.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_test_512.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/humanref_test_512_120f.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/manifest_summary.json +56 -0
- benchmark/samples/mixed_modalities_all_v1/text_test_1024.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/text_test_512_120f.jsonl +0 -0
- benchmark/samples/mixed_modalities_all_v1/text_transition_manifest_summary.json +42 -0
- benchmark/samples/mixed_modalities_all_v1/transition_manifest_summary.json +40 -0
README.md
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- motion-generation
|
| 5 |
+
- humanoid-robotics
|
| 6 |
+
- diffusion
|
| 7 |
+
- multimodal
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# OMG Release Artifacts
|
| 11 |
+
|
| 12 |
+
Official release checkpoints and benchmark artifacts for [OMG](https://github.com/Tsinghua-MARS-Lab/OMG).
|
| 13 |
+
|
| 14 |
+
## Checkpoints
|
| 15 |
+
|
| 16 |
+
| Model | Training step | File |
|
| 17 |
+
|---|---:|---|
|
| 18 |
+
| OMG 50M | 90,000 | `checkpoints/50m/sstep=090000.ckpt` |
|
| 19 |
+
| OMG 100M | 100,000 | `checkpoints/100m/sstep=100000.ckpt` |
|
| 20 |
+
| OMG 300M | 55,000 | `checkpoints/300m/sstep=055000.ckpt` |
|
| 21 |
+
| OMG 500M | 50,000 | `checkpoints/500m/sstep=050000.ckpt` |
|
| 22 |
+
|
| 23 |
+
The pretrained benchmark evaluator is stored at `evaluator/step_004000.pt`.
|
| 24 |
+
|
| 25 |
+
## Benchmark Artifacts
|
| 26 |
+
|
| 27 |
+
- `benchmark/samples/mixed_modalities_all_v1`: fixed benchmark inputs.
|
| 28 |
+
- `benchmark/results/mixed_modalities_all_v1`: generated motions, tracker-executed motions, and metric outputs.
|
| 29 |
+
- `SHA256SUMS`: checksums for release checkpoints and the evaluator.
|
| 30 |
+
|
| 31 |
+
See the [OMG benchmark documentation](https://github.com/Tsinghua-MARS-Lab/OMG/blob/main/docs/benchmark.md) for evaluation commands and [OMG-Data](https://huggingface.co/datasets/THU-MARS/OMG-Data) for training data.
|
| 32 |
+
|
SHA256SUMS
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
5e25dc7fd7251ec6df9dda8d3e008ff2540a5c8d1d691c576d4c64f9916eedaa checkpoints/100m/sstep=100000.ckpt
|
| 2 |
+
df4ce9aeced84fd96a39583c7f6e2c5773b76f942cb43012c6c51dd9eb806806 checkpoints/300m/sstep=055000.ckpt
|
| 3 |
+
f13bc58f07f104fcff67e35e79650def0fec5a6eb62a193fa1ae9f979fb911cf checkpoints/500m/sstep=050000.ckpt
|
| 4 |
+
389fae1aa2ba31234e0ef621706362a19387ac485dce1f929408e06a3b012b2c checkpoints/50m/sstep=090000.ckpt
|
| 5 |
+
e346790c5ce33c42840aa19f7fcc2891a4cab27e0d92e23bfaff7d48cb339c40 evaluator/step_004000.pt
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/aistpp_edge_metrics.json
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"div_g_gap": -0.13676034236105217,
|
| 3 |
+
"div_g_generated": 4.866377286829666,
|
| 4 |
+
"div_g_reference": 5.003137629190718,
|
| 5 |
+
"div_k_gap": -6.90296188275374,
|
| 6 |
+
"div_k_generated": 10.652930520737065,
|
| 7 |
+
"div_k_reference": 17.555892403490805,
|
| 8 |
+
"fid_g": 0.7416922948407674,
|
| 9 |
+
"fid_k": 59.19689074372258,
|
| 10 |
+
"frames": 60,
|
| 11 |
+
"geometric_feature": "g1_root_relative_body_geometry",
|
| 12 |
+
"geometric_feature_dim": 412,
|
| 13 |
+
"kinetic_feature": "fairmotion_root_relative_joint_kinetic_energy",
|
| 14 |
+
"kinetic_feature_dim": 90,
|
| 15 |
+
"left_foot_indices": [
|
| 16 |
+
5,
|
| 17 |
+
6
|
| 18 |
+
],
|
| 19 |
+
"num_samples": 512,
|
| 20 |
+
"pfc_gap": {
|
| 21 |
+
"max": 3.6477591193684584,
|
| 22 |
+
"mean": 0.0990505438447482,
|
| 23 |
+
"min": -4.8420685154923095,
|
| 24 |
+
"std": 0.8586369376448698
|
| 25 |
+
},
|
| 26 |
+
"pfc_generated": {
|
| 27 |
+
"max": 4.414909221516321,
|
| 28 |
+
"mean": 0.5746501914311459,
|
| 29 |
+
"min": 0.005071079598004716,
|
| 30 |
+
"std": 0.6583526046619211
|
| 31 |
+
},
|
| 32 |
+
"pfc_reference": {
|
| 33 |
+
"max": 5.707070070036645,
|
| 34 |
+
"mean": 0.4755996475863976,
|
| 35 |
+
"min": 1.734456963144955e-05,
|
| 36 |
+
"std": 0.7533267987077545
|
| 37 |
+
},
|
| 38 |
+
"right_foot_indices": [
|
| 39 |
+
11,
|
| 40 |
+
12
|
| 41 |
+
],
|
| 42 |
+
"root_index": 0,
|
| 43 |
+
"up_axis": "z"
|
| 44 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/artifact_samples.jsonl
ADDED
|
@@ -0,0 +1,512 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 3, "sample_index": 0}
|
| 2 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 7, "sample_index": 1}
|
| 3 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 12, "sample_index": 2}
|
| 4 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 23, "sample_index": 3}
|
| 5 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 32, "sample_index": 4}
|
| 6 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 41, "sample_index": 5}
|
| 7 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 48, "sample_index": 6}
|
| 8 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 58, "sample_index": 7}
|
| 9 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 73, "sample_index": 8}
|
| 10 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 106, "sample_index": 9}
|
| 11 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 108, "sample_index": 10}
|
| 12 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 116, "sample_index": 11}
|
| 13 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 127, "sample_index": 12}
|
| 14 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 129, "sample_index": 13}
|
| 15 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 181, "sample_index": 14}
|
| 16 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 200, "sample_index": 15}
|
| 17 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 249, "sample_index": 16}
|
| 18 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 254, "sample_index": 17}
|
| 19 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 342, "sample_index": 18}
|
| 20 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 376, "sample_index": 19}
|
| 21 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 381, "sample_index": 20}
|
| 22 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 382, "sample_index": 21}
|
| 23 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 398, "sample_index": 22}
|
| 24 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 435, "sample_index": 23}
|
| 25 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 437, "sample_index": 24}
|
| 26 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 461, "sample_index": 25}
|
| 27 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 480, "sample_index": 26}
|
| 28 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 482, "sample_index": 27}
|
| 29 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 494, "sample_index": 28}
|
| 30 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 507, "sample_index": 29}
|
| 31 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 534, "sample_index": 30}
|
| 32 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 557, "sample_index": 31}
|
| 33 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 558, "sample_index": 32}
|
| 34 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 562, "sample_index": 33}
|
| 35 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 567, "sample_index": 34}
|
| 36 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 570, "sample_index": 35}
|
| 37 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 575, "sample_index": 36}
|
| 38 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 587, "sample_index": 37}
|
| 39 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 589, "sample_index": 38}
|
| 40 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 607, "sample_index": 39}
|
| 41 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 615, "sample_index": 40}
|
| 42 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 629, "sample_index": 41}
|
| 43 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 676, "sample_index": 42}
|
| 44 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 699, "sample_index": 43}
|
| 45 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 719, "sample_index": 44}
|
| 46 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 723, "sample_index": 45}
|
| 47 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 724, "sample_index": 46}
|
| 48 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 757, "sample_index": 47}
|
| 49 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 768, "sample_index": 48}
|
| 50 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 779, "sample_index": 49}
|
| 51 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 780, "sample_index": 50}
|
| 52 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 785, "sample_index": 51}
|
| 53 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 793, "sample_index": 52}
|
| 54 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 799, "sample_index": 53}
|
| 55 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 802, "sample_index": 54}
|
| 56 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 855, "sample_index": 55}
|
| 57 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 857, "sample_index": 56}
|
| 58 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 866, "sample_index": 57}
|
| 59 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 892, "sample_index": 58}
|
| 60 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 901, "sample_index": 59}
|
| 61 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 902, "sample_index": 60}
|
| 62 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 905, "sample_index": 61}
|
| 63 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 926, "sample_index": 62}
|
| 64 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 941, "sample_index": 63}
|
| 65 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 946, "sample_index": 64}
|
| 66 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 959, "sample_index": 65}
|
| 67 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 964, "sample_index": 66}
|
| 68 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 979, "sample_index": 67}
|
| 69 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1006, "sample_index": 68}
|
| 70 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1009, "sample_index": 69}
|
| 71 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1016, "sample_index": 70}
|
| 72 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1040, "sample_index": 71}
|
| 73 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1043, "sample_index": 72}
|
| 74 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1054, "sample_index": 73}
|
| 75 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1069, "sample_index": 74}
|
| 76 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1071, "sample_index": 75}
|
| 77 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1078, "sample_index": 76}
|
| 78 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1079, "sample_index": 77}
|
| 79 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1104, "sample_index": 78}
|
| 80 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1120, "sample_index": 79}
|
| 81 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1144, "sample_index": 80}
|
| 82 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1158, "sample_index": 81}
|
| 83 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1172, "sample_index": 82}
|
| 84 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1183, "sample_index": 83}
|
| 85 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1194, "sample_index": 84}
|
| 86 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1204, "sample_index": 85}
|
| 87 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1224, "sample_index": 86}
|
| 88 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1235, "sample_index": 87}
|
| 89 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1240, "sample_index": 88}
|
| 90 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1250, "sample_index": 89}
|
| 91 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1254, "sample_index": 90}
|
| 92 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1261, "sample_index": 91}
|
| 93 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1295, "sample_index": 92}
|
| 94 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1302, "sample_index": 93}
|
| 95 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1326, "sample_index": 94}
|
| 96 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1341, "sample_index": 95}
|
| 97 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1388, "sample_index": 96}
|
| 98 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1394, "sample_index": 97}
|
| 99 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1400, "sample_index": 98}
|
| 100 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1442, "sample_index": 99}
|
| 101 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1464, "sample_index": 100}
|
| 102 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1506, "sample_index": 101}
|
| 103 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "dataset_index": 1507, "sample_index": 102}
|
| 104 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 19, "sample_index": 103}
|
| 105 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 30, "sample_index": 104}
|
| 106 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 50, "sample_index": 105}
|
| 107 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 53, "sample_index": 106}
|
| 108 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 76, "sample_index": 107}
|
| 109 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 82, "sample_index": 108}
|
| 110 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 84, "sample_index": 109}
|
| 111 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 118, "sample_index": 110}
|
| 112 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 126, "sample_index": 111}
|
| 113 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 127, "sample_index": 112}
|
| 114 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 156, "sample_index": 113}
|
| 115 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 160, "sample_index": 114}
|
| 116 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 174, "sample_index": 115}
|
| 117 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 202, "sample_index": 116}
|
| 118 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 217, "sample_index": 117}
|
| 119 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 224, "sample_index": 118}
|
| 120 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 241, "sample_index": 119}
|
| 121 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 242, "sample_index": 120}
|
| 122 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 263, "sample_index": 121}
|
| 123 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 271, "sample_index": 122}
|
| 124 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 281, "sample_index": 123}
|
| 125 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 285, "sample_index": 124}
|
| 126 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 286, "sample_index": 125}
|
| 127 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 348, "sample_index": 126}
|
| 128 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 404, "sample_index": 127}
|
| 129 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 422, "sample_index": 128}
|
| 130 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 434, "sample_index": 129}
|
| 131 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 436, "sample_index": 130}
|
| 132 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 453, "sample_index": 131}
|
| 133 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 466, "sample_index": 132}
|
| 134 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 499, "sample_index": 133}
|
| 135 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 502, "sample_index": 134}
|
| 136 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 512, "sample_index": 135}
|
| 137 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 513, "sample_index": 136}
|
| 138 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 525, "sample_index": 137}
|
| 139 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 558, "sample_index": 138}
|
| 140 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 561, "sample_index": 139}
|
| 141 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 590, "sample_index": 140}
|
| 142 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 661, "sample_index": 141}
|
| 143 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 723, "sample_index": 142}
|
| 144 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 800, "sample_index": 143}
|
| 145 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 805, "sample_index": 144}
|
| 146 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 820, "sample_index": 145}
|
| 147 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 827, "sample_index": 146}
|
| 148 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 837, "sample_index": 147}
|
| 149 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 881, "sample_index": 148}
|
| 150 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 885, "sample_index": 149}
|
| 151 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 947, "sample_index": 150}
|
| 152 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 953, "sample_index": 151}
|
| 153 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 959, "sample_index": 152}
|
| 154 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 979, "sample_index": 153}
|
| 155 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1018, "sample_index": 154}
|
| 156 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1076, "sample_index": 155}
|
| 157 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1082, "sample_index": 156}
|
| 158 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1083, "sample_index": 157}
|
| 159 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1090, "sample_index": 158}
|
| 160 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1137, "sample_index": 159}
|
| 161 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1165, "sample_index": 160}
|
| 162 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1167, "sample_index": 161}
|
| 163 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1170, "sample_index": 162}
|
| 164 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1197, "sample_index": 163}
|
| 165 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1224, "sample_index": 164}
|
| 166 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1242, "sample_index": 165}
|
| 167 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1246, "sample_index": 166}
|
| 168 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1247, "sample_index": 167}
|
| 169 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1296, "sample_index": 168}
|
| 170 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1311, "sample_index": 169}
|
| 171 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1369, "sample_index": 170}
|
| 172 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1388, "sample_index": 171}
|
| 173 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1389, "sample_index": 172}
|
| 174 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1446, "sample_index": 173}
|
| 175 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1467, "sample_index": 174}
|
| 176 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1484, "sample_index": 175}
|
| 177 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1507, "sample_index": 176}
|
| 178 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1510, "sample_index": 177}
|
| 179 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1544, "sample_index": 178}
|
| 180 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1555, "sample_index": 179}
|
| 181 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1559, "sample_index": 180}
|
| 182 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1574, "sample_index": 181}
|
| 183 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1596, "sample_index": 182}
|
| 184 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1611, "sample_index": 183}
|
| 185 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1612, "sample_index": 184}
|
| 186 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1624, "sample_index": 185}
|
| 187 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1641, "sample_index": 186}
|
| 188 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1666, "sample_index": 187}
|
| 189 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1679, "sample_index": 188}
|
| 190 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1700, "sample_index": 189}
|
| 191 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1715, "sample_index": 190}
|
| 192 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1717, "sample_index": 191}
|
| 193 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1741, "sample_index": 192}
|
| 194 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1763, "sample_index": 193}
|
| 195 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1764, "sample_index": 194}
|
| 196 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1776, "sample_index": 195}
|
| 197 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1792, "sample_index": 196}
|
| 198 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1832, "sample_index": 197}
|
| 199 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1854, "sample_index": 198}
|
| 200 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1856, "sample_index": 199}
|
| 201 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1858, "sample_index": 200}
|
| 202 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1875, "sample_index": 201}
|
| 203 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1949, "sample_index": 202}
|
| 204 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1963, "sample_index": 203}
|
| 205 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1966, "sample_index": 204}
|
| 206 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "dataset_index": 1981, "sample_index": 205}
|
| 207 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 6, "sample_index": 206}
|
| 208 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 26, "sample_index": 207}
|
| 209 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 42, "sample_index": 208}
|
| 210 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 47, "sample_index": 209}
|
| 211 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 60, "sample_index": 210}
|
| 212 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 64, "sample_index": 211}
|
| 213 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 71, "sample_index": 212}
|
| 214 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 73, "sample_index": 213}
|
| 215 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 76, "sample_index": 214}
|
| 216 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 88, "sample_index": 215}
|
| 217 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 89, "sample_index": 216}
|
| 218 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 92, "sample_index": 217}
|
| 219 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 98, "sample_index": 218}
|
| 220 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 118, "sample_index": 219}
|
| 221 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 128, "sample_index": 220}
|
| 222 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 131, "sample_index": 221}
|
| 223 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 134, "sample_index": 222}
|
| 224 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 139, "sample_index": 223}
|
| 225 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 170, "sample_index": 224}
|
| 226 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 180, "sample_index": 225}
|
| 227 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 196, "sample_index": 226}
|
| 228 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 210, "sample_index": 227}
|
| 229 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 219, "sample_index": 228}
|
| 230 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 237, "sample_index": 229}
|
| 231 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 238, "sample_index": 230}
|
| 232 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 272, "sample_index": 231}
|
| 233 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 274, "sample_index": 232}
|
| 234 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 276, "sample_index": 233}
|
| 235 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 280, "sample_index": 234}
|
| 236 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 286, "sample_index": 235}
|
| 237 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 318, "sample_index": 236}
|
| 238 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 324, "sample_index": 237}
|
| 239 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 333, "sample_index": 238}
|
| 240 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 356, "sample_index": 239}
|
| 241 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 375, "sample_index": 240}
|
| 242 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 393, "sample_index": 241}
|
| 243 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 399, "sample_index": 242}
|
| 244 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 404, "sample_index": 243}
|
| 245 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 417, "sample_index": 244}
|
| 246 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 465, "sample_index": 245}
|
| 247 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 475, "sample_index": 246}
|
| 248 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 484, "sample_index": 247}
|
| 249 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 503, "sample_index": 248}
|
| 250 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 505, "sample_index": 249}
|
| 251 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 517, "sample_index": 250}
|
| 252 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 518, "sample_index": 251}
|
| 253 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 521, "sample_index": 252}
|
| 254 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 526, "sample_index": 253}
|
| 255 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 534, "sample_index": 254}
|
| 256 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 536, "sample_index": 255}
|
| 257 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 539, "sample_index": 256}
|
| 258 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 556, "sample_index": 257}
|
| 259 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 559, "sample_index": 258}
|
| 260 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 560, "sample_index": 259}
|
| 261 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 595, "sample_index": 260}
|
| 262 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 603, "sample_index": 261}
|
| 263 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 614, "sample_index": 262}
|
| 264 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 615, "sample_index": 263}
|
| 265 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 652, "sample_index": 264}
|
| 266 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 660, "sample_index": 265}
|
| 267 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 688, "sample_index": 266}
|
| 268 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 696, "sample_index": 267}
|
| 269 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 699, "sample_index": 268}
|
| 270 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 705, "sample_index": 269}
|
| 271 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 710, "sample_index": 270}
|
| 272 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 721, "sample_index": 271}
|
| 273 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 726, "sample_index": 272}
|
| 274 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 738, "sample_index": 273}
|
| 275 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 747, "sample_index": 274}
|
| 276 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 751, "sample_index": 275}
|
| 277 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 761, "sample_index": 276}
|
| 278 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 782, "sample_index": 277}
|
| 279 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 785, "sample_index": 278}
|
| 280 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 788, "sample_index": 279}
|
| 281 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 826, "sample_index": 280}
|
| 282 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 834, "sample_index": 281}
|
| 283 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 857, "sample_index": 282}
|
| 284 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 860, "sample_index": 283}
|
| 285 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 861, "sample_index": 284}
|
| 286 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 877, "sample_index": 285}
|
| 287 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 885, "sample_index": 286}
|
| 288 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 893, "sample_index": 287}
|
| 289 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 898, "sample_index": 288}
|
| 290 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 902, "sample_index": 289}
|
| 291 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 910, "sample_index": 290}
|
| 292 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 920, "sample_index": 291}
|
| 293 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 937, "sample_index": 292}
|
| 294 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 943, "sample_index": 293}
|
| 295 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 944, "sample_index": 294}
|
| 296 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 953, "sample_index": 295}
|
| 297 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 959, "sample_index": 296}
|
| 298 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 965, "sample_index": 297}
|
| 299 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 991, "sample_index": 298}
|
| 300 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1001, "sample_index": 299}
|
| 301 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1017, "sample_index": 300}
|
| 302 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1026, "sample_index": 301}
|
| 303 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1033, "sample_index": 302}
|
| 304 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1036, "sample_index": 303}
|
| 305 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1052, "sample_index": 304}
|
| 306 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1056, "sample_index": 305}
|
| 307 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1060, "sample_index": 306}
|
| 308 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "dataset_index": 1088, "sample_index": 307}
|
| 309 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 21, "sample_index": 308}
|
| 310 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 41, "sample_index": 309}
|
| 311 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 58, "sample_index": 310}
|
| 312 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 118, "sample_index": 311}
|
| 313 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 128, "sample_index": 312}
|
| 314 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 145, "sample_index": 313}
|
| 315 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 210, "sample_index": 314}
|
| 316 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 214, "sample_index": 315}
|
| 317 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 233, "sample_index": 316}
|
| 318 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 240, "sample_index": 317}
|
| 319 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 263, "sample_index": 318}
|
| 320 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 266, "sample_index": 319}
|
| 321 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 280, "sample_index": 320}
|
| 322 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 285, "sample_index": 321}
|
| 323 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 295, "sample_index": 322}
|
| 324 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 298, "sample_index": 323}
|
| 325 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 301, "sample_index": 324}
|
| 326 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 303, "sample_index": 325}
|
| 327 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 321, "sample_index": 326}
|
| 328 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 333, "sample_index": 327}
|
| 329 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 339, "sample_index": 328}
|
| 330 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 341, "sample_index": 329}
|
| 331 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 350, "sample_index": 330}
|
| 332 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 384, "sample_index": 331}
|
| 333 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 387, "sample_index": 332}
|
| 334 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 423, "sample_index": 333}
|
| 335 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 429, "sample_index": 334}
|
| 336 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 467, "sample_index": 335}
|
| 337 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 471, "sample_index": 336}
|
| 338 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 489, "sample_index": 337}
|
| 339 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 522, "sample_index": 338}
|
| 340 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 535, "sample_index": 339}
|
| 341 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 538, "sample_index": 340}
|
| 342 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 547, "sample_index": 341}
|
| 343 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 550, "sample_index": 342}
|
| 344 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 557, "sample_index": 343}
|
| 345 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 563, "sample_index": 344}
|
| 346 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 585, "sample_index": 345}
|
| 347 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 619, "sample_index": 346}
|
| 348 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 625, "sample_index": 347}
|
| 349 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 628, "sample_index": 348}
|
| 350 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 655, "sample_index": 349}
|
| 351 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 688, "sample_index": 350}
|
| 352 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 705, "sample_index": 351}
|
| 353 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 723, "sample_index": 352}
|
| 354 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 726, "sample_index": 353}
|
| 355 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 736, "sample_index": 354}
|
| 356 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 820, "sample_index": 355}
|
| 357 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 869, "sample_index": 356}
|
| 358 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 878, "sample_index": 357}
|
| 359 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 884, "sample_index": 358}
|
| 360 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 953, "sample_index": 359}
|
| 361 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 954, "sample_index": 360}
|
| 362 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 960, "sample_index": 361}
|
| 363 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 964, "sample_index": 362}
|
| 364 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 987, "sample_index": 363}
|
| 365 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 996, "sample_index": 364}
|
| 366 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1002, "sample_index": 365}
|
| 367 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1032, "sample_index": 366}
|
| 368 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1033, "sample_index": 367}
|
| 369 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1035, "sample_index": 368}
|
| 370 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1063, "sample_index": 369}
|
| 371 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1065, "sample_index": 370}
|
| 372 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1068, "sample_index": 371}
|
| 373 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1083, "sample_index": 372}
|
| 374 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1100, "sample_index": 373}
|
| 375 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1104, "sample_index": 374}
|
| 376 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1107, "sample_index": 375}
|
| 377 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1122, "sample_index": 376}
|
| 378 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1150, "sample_index": 377}
|
| 379 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1151, "sample_index": 378}
|
| 380 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1182, "sample_index": 379}
|
| 381 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1215, "sample_index": 380}
|
| 382 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1221, "sample_index": 381}
|
| 383 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1222, "sample_index": 382}
|
| 384 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1223, "sample_index": 383}
|
| 385 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1235, "sample_index": 384}
|
| 386 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1243, "sample_index": 385}
|
| 387 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1261, "sample_index": 386}
|
| 388 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1262, "sample_index": 387}
|
| 389 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1289, "sample_index": 388}
|
| 390 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1297, "sample_index": 389}
|
| 391 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1321, "sample_index": 390}
|
| 392 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1326, "sample_index": 391}
|
| 393 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1342, "sample_index": 392}
|
| 394 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1350, "sample_index": 393}
|
| 395 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1354, "sample_index": 394}
|
| 396 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1359, "sample_index": 395}
|
| 397 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1457, "sample_index": 396}
|
| 398 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1520, "sample_index": 397}
|
| 399 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1523, "sample_index": 398}
|
| 400 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1579, "sample_index": 399}
|
| 401 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1593, "sample_index": 400}
|
| 402 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1614, "sample_index": 401}
|
| 403 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1617, "sample_index": 402}
|
| 404 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1649, "sample_index": 403}
|
| 405 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1662, "sample_index": 404}
|
| 406 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1668, "sample_index": 405}
|
| 407 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1684, "sample_index": 406}
|
| 408 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1694, "sample_index": 407}
|
| 409 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1735, "sample_index": 408}
|
| 410 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "dataset_index": 1738, "sample_index": 409}
|
| 411 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 14, "sample_index": 410}
|
| 412 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 15, "sample_index": 411}
|
| 413 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 38, "sample_index": 412}
|
| 414 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 49, "sample_index": 413}
|
| 415 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 57, "sample_index": 414}
|
| 416 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 58, "sample_index": 415}
|
| 417 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 73, "sample_index": 416}
|
| 418 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 86, "sample_index": 417}
|
| 419 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 89, "sample_index": 418}
|
| 420 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 90, "sample_index": 419}
|
| 421 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 94, "sample_index": 420}
|
| 422 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 98, "sample_index": 421}
|
| 423 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 99, "sample_index": 422}
|
| 424 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 110, "sample_index": 423}
|
| 425 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 111, "sample_index": 424}
|
| 426 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 113, "sample_index": 425}
|
| 427 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 118, "sample_index": 426}
|
| 428 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 127, "sample_index": 427}
|
| 429 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 140, "sample_index": 428}
|
| 430 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 142, "sample_index": 429}
|
| 431 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 149, "sample_index": 430}
|
| 432 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 158, "sample_index": 431}
|
| 433 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 159, "sample_index": 432}
|
| 434 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 167, "sample_index": 433}
|
| 435 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 173, "sample_index": 434}
|
| 436 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 175, "sample_index": 435}
|
| 437 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 183, "sample_index": 436}
|
| 438 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 184, "sample_index": 437}
|
| 439 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 189, "sample_index": 438}
|
| 440 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 210, "sample_index": 439}
|
| 441 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 216, "sample_index": 440}
|
| 442 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 217, "sample_index": 441}
|
| 443 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 223, "sample_index": 442}
|
| 444 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 230, "sample_index": 443}
|
| 445 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 245, "sample_index": 444}
|
| 446 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 255, "sample_index": 445}
|
| 447 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 263, "sample_index": 446}
|
| 448 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 266, "sample_index": 447}
|
| 449 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 271, "sample_index": 448}
|
| 450 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 273, "sample_index": 449}
|
| 451 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 275, "sample_index": 450}
|
| 452 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 278, "sample_index": 451}
|
| 453 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 281, "sample_index": 452}
|
| 454 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 284, "sample_index": 453}
|
| 455 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 286, "sample_index": 454}
|
| 456 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 291, "sample_index": 455}
|
| 457 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 292, "sample_index": 456}
|
| 458 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 302, "sample_index": 457}
|
| 459 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 305, "sample_index": 458}
|
| 460 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 306, "sample_index": 459}
|
| 461 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 320, "sample_index": 460}
|
| 462 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 326, "sample_index": 461}
|
| 463 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 332, "sample_index": 462}
|
| 464 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 333, "sample_index": 463}
|
| 465 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 336, "sample_index": 464}
|
| 466 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 339, "sample_index": 465}
|
| 467 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 340, "sample_index": 466}
|
| 468 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 347, "sample_index": 467}
|
| 469 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 358, "sample_index": 468}
|
| 470 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 360, "sample_index": 469}
|
| 471 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 378, "sample_index": 470}
|
| 472 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 380, "sample_index": 471}
|
| 473 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 388, "sample_index": 472}
|
| 474 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 390, "sample_index": 473}
|
| 475 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 396, "sample_index": 474}
|
| 476 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 399, "sample_index": 475}
|
| 477 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 402, "sample_index": 476}
|
| 478 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 404, "sample_index": 477}
|
| 479 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 406, "sample_index": 478}
|
| 480 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 408, "sample_index": 479}
|
| 481 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 413, "sample_index": 480}
|
| 482 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 418, "sample_index": 481}
|
| 483 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 427, "sample_index": 482}
|
| 484 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 432, "sample_index": 483}
|
| 485 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 433, "sample_index": 484}
|
| 486 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 439, "sample_index": 485}
|
| 487 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 446, "sample_index": 486}
|
| 488 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 447, "sample_index": 487}
|
| 489 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 454, "sample_index": 488}
|
| 490 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 457, "sample_index": 489}
|
| 491 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 468, "sample_index": 490}
|
| 492 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 480, "sample_index": 491}
|
| 493 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 492, "sample_index": 492}
|
| 494 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 495, "sample_index": 493}
|
| 495 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 499, "sample_index": 494}
|
| 496 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 500, "sample_index": 495}
|
| 497 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 506, "sample_index": 496}
|
| 498 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 515, "sample_index": 497}
|
| 499 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 549, "sample_index": 498}
|
| 500 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 551, "sample_index": 499}
|
| 501 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 553, "sample_index": 500}
|
| 502 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 555, "sample_index": 501}
|
| 503 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 558, "sample_index": 502}
|
| 504 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 559, "sample_index": 503}
|
| 505 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 567, "sample_index": 504}
|
| 506 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 569, "sample_index": 505}
|
| 507 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 570, "sample_index": 506}
|
| 508 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 585, "sample_index": 507}
|
| 509 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 588, "sample_index": 508}
|
| 510 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 592, "sample_index": 509}
|
| 511 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 594, "sample_index": 510}
|
| 512 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "dataset_index": 596, "sample_index": 511}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/audio_metrics.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"beat_align_gap": {
|
| 3 |
+
"max": 0.5701364518161618,
|
| 4 |
+
"mean": -0.06749126869399916,
|
| 5 |
+
"min": -0.7608427161001056,
|
| 6 |
+
"std": 0.18871442298460653
|
| 7 |
+
},
|
| 8 |
+
"beat_align_generated": {
|
| 9 |
+
"max": 0.9819864896355885,
|
| 10 |
+
"mean": 0.49377759948027955,
|
| 11 |
+
"min": 0.0,
|
| 12 |
+
"std": 0.14606512824675702
|
| 13 |
+
},
|
| 14 |
+
"beat_align_reference": {
|
| 15 |
+
"max": 0.9864898672266914,
|
| 16 |
+
"mean": 0.5612688681742788,
|
| 17 |
+
"min": 0.0,
|
| 18 |
+
"std": 0.13609287544141188
|
| 19 |
+
},
|
| 20 |
+
"beat_direction": "music_to_motion",
|
| 21 |
+
"beat_threshold": 0.5,
|
| 22 |
+
"frames": 60,
|
| 23 |
+
"min_motion_beat_distance_seconds": 0.25,
|
| 24 |
+
"num_samples": 512,
|
| 25 |
+
"sigma_frames": 3.0
|
| 26 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/benchmark.json
ADDED
|
@@ -0,0 +1,1058 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"aistpp_edge_metrics": {
|
| 3 |
+
"div_g_gap": -0.13676034236105217,
|
| 4 |
+
"div_g_generated": 4.866377286829666,
|
| 5 |
+
"div_g_reference": 5.003137629190718,
|
| 6 |
+
"div_k_gap": -6.90296188275374,
|
| 7 |
+
"div_k_generated": 10.652930520737065,
|
| 8 |
+
"div_k_reference": 17.555892403490805,
|
| 9 |
+
"fid_g": 0.7416922948407674,
|
| 10 |
+
"fid_k": 59.19689074372258,
|
| 11 |
+
"frames": 60,
|
| 12 |
+
"geometric_feature": "g1_root_relative_body_geometry",
|
| 13 |
+
"geometric_feature_dim": 412,
|
| 14 |
+
"kinetic_feature": "fairmotion_root_relative_joint_kinetic_energy",
|
| 15 |
+
"kinetic_feature_dim": 90,
|
| 16 |
+
"left_foot_indices": [
|
| 17 |
+
5,
|
| 18 |
+
6
|
| 19 |
+
],
|
| 20 |
+
"num_samples": 512,
|
| 21 |
+
"pfc_gap": {
|
| 22 |
+
"max": 3.6477591193684584,
|
| 23 |
+
"mean": 0.0990505438447482,
|
| 24 |
+
"min": -4.8420685154923095,
|
| 25 |
+
"std": 0.8586369376448698
|
| 26 |
+
},
|
| 27 |
+
"pfc_generated": {
|
| 28 |
+
"max": 4.414909221516321,
|
| 29 |
+
"mean": 0.5746501914311459,
|
| 30 |
+
"min": 0.005071079598004716,
|
| 31 |
+
"std": 0.6583526046619211
|
| 32 |
+
},
|
| 33 |
+
"pfc_reference": {
|
| 34 |
+
"max": 5.707070070036645,
|
| 35 |
+
"mean": 0.4755996475863976,
|
| 36 |
+
"min": 1.734456963144955e-05,
|
| 37 |
+
"std": 0.7533267987077545
|
| 38 |
+
},
|
| 39 |
+
"right_foot_indices": [
|
| 40 |
+
11,
|
| 41 |
+
12
|
| 42 |
+
],
|
| 43 |
+
"root_index": 0,
|
| 44 |
+
"up_axis": "z"
|
| 45 |
+
},
|
| 46 |
+
"audio_metrics": {
|
| 47 |
+
"beat_align_gap": {
|
| 48 |
+
"max": 0.5701364518161618,
|
| 49 |
+
"mean": -0.06749126869399916,
|
| 50 |
+
"min": -0.7608427161001056,
|
| 51 |
+
"std": 0.18871442298460653
|
| 52 |
+
},
|
| 53 |
+
"beat_align_generated": {
|
| 54 |
+
"max": 0.9819864896355885,
|
| 55 |
+
"mean": 0.49377759948027955,
|
| 56 |
+
"min": 0.0,
|
| 57 |
+
"std": 0.14606512824675702
|
| 58 |
+
},
|
| 59 |
+
"beat_align_reference": {
|
| 60 |
+
"max": 0.9864898672266914,
|
| 61 |
+
"mean": 0.5612688681742788,
|
| 62 |
+
"min": 0.0,
|
| 63 |
+
"std": 0.13609287544141188
|
| 64 |
+
},
|
| 65 |
+
"beat_direction": "music_to_motion",
|
| 66 |
+
"beat_threshold": 0.5,
|
| 67 |
+
"frames": 60,
|
| 68 |
+
"min_motion_beat_distance_seconds": 0.25,
|
| 69 |
+
"num_samples": 512,
|
| 70 |
+
"sigma_frames": 3.0
|
| 71 |
+
},
|
| 72 |
+
"benchmark": "artifact_audio",
|
| 73 |
+
"dataset_metrics": {
|
| 74 |
+
"aioz_gdance_original_audio_test": {
|
| 75 |
+
"aistpp_edge": {
|
| 76 |
+
"div_g_gap": 0.23862403086776185,
|
| 77 |
+
"div_g_generated": 3.7473771961157687,
|
| 78 |
+
"div_g_reference": 3.508753165248007,
|
| 79 |
+
"div_k_gap": -1.4359521145663674,
|
| 80 |
+
"div_k_generated": 9.315809033822825,
|
| 81 |
+
"div_k_reference": 10.751761148389193,
|
| 82 |
+
"fid_g": 1.4318985574865248,
|
| 83 |
+
"fid_k": 7.601990475458294,
|
| 84 |
+
"geometric_feature_dim": 412,
|
| 85 |
+
"kinetic_feature_dim": 90,
|
| 86 |
+
"num_samples": 103,
|
| 87 |
+
"pfc_gap": {
|
| 88 |
+
"max": 3.2058528208797874,
|
| 89 |
+
"mean": 0.05262750505804416,
|
| 90 |
+
"min": -1.6349628441701682,
|
| 91 |
+
"std": 0.6426666823502454
|
| 92 |
+
},
|
| 93 |
+
"pfc_generated": {
|
| 94 |
+
"max": 3.4774407560932494,
|
| 95 |
+
"mean": 0.3836619604251987,
|
| 96 |
+
"min": 0.005071079598004716,
|
| 97 |
+
"std": 0.5183719007045057
|
| 98 |
+
},
|
| 99 |
+
"pfc_reference": {
|
| 100 |
+
"max": 1.8273494537376207,
|
| 101 |
+
"mean": 0.3310344553671544,
|
| 102 |
+
"min": 0.0002850814965562205,
|
| 103 |
+
"std": 0.42521448548204027
|
| 104 |
+
}
|
| 105 |
+
},
|
| 106 |
+
"audio": {
|
| 107 |
+
"beat_align_gap": {
|
| 108 |
+
"max": 0.23840448764256933,
|
| 109 |
+
"mean": -0.033628672950055886,
|
| 110 |
+
"min": -0.32453708581516766,
|
| 111 |
+
"std": 0.13742288780504877
|
| 112 |
+
},
|
| 113 |
+
"beat_align_generated": {
|
| 114 |
+
"max": 0.6725372901371526,
|
| 115 |
+
"mean": 0.4919088098323277,
|
| 116 |
+
"min": 0.24963270016294453,
|
| 117 |
+
"std": 0.09099816858378701
|
| 118 |
+
},
|
| 119 |
+
"beat_align_reference": {
|
| 120 |
+
"max": 0.723284948504687,
|
| 121 |
+
"mean": 0.5255374827823837,
|
| 122 |
+
"min": 0.33650493752041727,
|
| 123 |
+
"std": 0.08282468265444852
|
| 124 |
+
},
|
| 125 |
+
"num_samples": 103
|
| 126 |
+
},
|
| 127 |
+
"num_samples": 103,
|
| 128 |
+
"physical": {
|
| 129 |
+
"body_jerk_mean": {
|
| 130 |
+
"max": 187.8517608642578,
|
| 131 |
+
"mean": 75.8366039979805,
|
| 132 |
+
"min": 27.45358657836914,
|
| 133 |
+
"std": 24.36331538169348
|
| 134 |
+
},
|
| 135 |
+
"contact_sliding_speed": {
|
| 136 |
+
"max": 0.8738095164299011,
|
| 137 |
+
"mean": 0.3195253503915754,
|
| 138 |
+
"min": 0.043544918298721313,
|
| 139 |
+
"std": 0.16198516039471123
|
| 140 |
+
},
|
| 141 |
+
"foot_ground_error": {
|
| 142 |
+
"max": 0.08582238852977753,
|
| 143 |
+
"mean": 0.04983966779054224,
|
| 144 |
+
"min": 0.007044824305921793,
|
| 145 |
+
"std": 0.017409112859072134
|
| 146 |
+
},
|
| 147 |
+
"num_samples": 103
|
| 148 |
+
}
|
| 149 |
+
},
|
| 150 |
+
"aistpp_original_audio_test": {
|
| 151 |
+
"aistpp_edge": {
|
| 152 |
+
"div_g_gap": 0.3749239056934046,
|
| 153 |
+
"div_g_generated": 3.929406542818343,
|
| 154 |
+
"div_g_reference": 3.5544826371249383,
|
| 155 |
+
"div_k_gap": -12.850860561893567,
|
| 156 |
+
"div_k_generated": 12.909821306156998,
|
| 157 |
+
"div_k_reference": 25.760681868050565,
|
| 158 |
+
"fid_g": 2.601341460428607,
|
| 159 |
+
"fid_k": 251.18791168722862,
|
| 160 |
+
"geometric_feature_dim": 412,
|
| 161 |
+
"kinetic_feature_dim": 90,
|
| 162 |
+
"num_samples": 103,
|
| 163 |
+
"pfc_gap": {
|
| 164 |
+
"max": 2.477920920125234,
|
| 165 |
+
"mean": -0.24784057237073714,
|
| 166 |
+
"min": -4.8420685154923095,
|
| 167 |
+
"std": 1.3450989582965898
|
| 168 |
+
},
|
| 169 |
+
"pfc_generated": {
|
| 170 |
+
"max": 3.0173064117838173,
|
| 171 |
+
"mean": 0.8859167056460855,
|
| 172 |
+
"min": 0.03606813460559294,
|
| 173 |
+
"std": 0.6825868113283983
|
| 174 |
+
},
|
| 175 |
+
"pfc_reference": {
|
| 176 |
+
"max": 5.707070070036645,
|
| 177 |
+
"mean": 1.1337572780168226,
|
| 178 |
+
"min": 7.906044911863323e-05,
|
| 179 |
+
"std": 1.2009778419653923
|
| 180 |
+
}
|
| 181 |
+
},
|
| 182 |
+
"audio": {
|
| 183 |
+
"beat_align_gap": {
|
| 184 |
+
"max": 0.5701364518161618,
|
| 185 |
+
"mean": -0.0801368416977987,
|
| 186 |
+
"min": -0.7608427161001056,
|
| 187 |
+
"std": 0.28185526909472197
|
| 188 |
+
},
|
| 189 |
+
"beat_align_generated": {
|
| 190 |
+
"max": 0.9819864896355885,
|
| 191 |
+
"mean": 0.5061291397536287,
|
| 192 |
+
"min": 0.0,
|
| 193 |
+
"std": 0.2064175413162069
|
| 194 |
+
},
|
| 195 |
+
"beat_align_reference": {
|
| 196 |
+
"max": 0.9864898672266914,
|
| 197 |
+
"mean": 0.5862659814514274,
|
| 198 |
+
"min": 0.0,
|
| 199 |
+
"std": 0.19459278456334647
|
| 200 |
+
},
|
| 201 |
+
"num_samples": 103
|
| 202 |
+
},
|
| 203 |
+
"num_samples": 103,
|
| 204 |
+
"physical": {
|
| 205 |
+
"body_jerk_mean": {
|
| 206 |
+
"max": 278.0105285644531,
|
| 207 |
+
"mean": 135.99853585993202,
|
| 208 |
+
"min": 32.689208984375,
|
| 209 |
+
"std": 46.387850301317464
|
| 210 |
+
},
|
| 211 |
+
"contact_sliding_speed": {
|
| 212 |
+
"max": 1.3322360515594482,
|
| 213 |
+
"mean": 0.5159466229017499,
|
| 214 |
+
"min": 0.0,
|
| 215 |
+
"std": 0.24116125404927893
|
| 216 |
+
},
|
| 217 |
+
"foot_ground_error": {
|
| 218 |
+
"max": 0.18515852093696594,
|
| 219 |
+
"mean": 0.030137372799941057,
|
| 220 |
+
"min": 0.004513147287070751,
|
| 221 |
+
"std": 0.028280388956618472
|
| 222 |
+
},
|
| 223 |
+
"num_samples": 103
|
| 224 |
+
}
|
| 225 |
+
},
|
| 226 |
+
"compas3d_original_audio_test": {
|
| 227 |
+
"aistpp_edge": {
|
| 228 |
+
"div_g_gap": 0.17526199534596376,
|
| 229 |
+
"div_g_generated": 4.703408486137112,
|
| 230 |
+
"div_g_reference": 4.528146490791149,
|
| 231 |
+
"div_k_gap": -3.1528614038078047,
|
| 232 |
+
"div_k_generated": 8.431761165269869,
|
| 233 |
+
"div_k_reference": 11.584622569077673,
|
| 234 |
+
"fid_g": 2.805483949908389,
|
| 235 |
+
"fid_k": 67.40567433330911,
|
| 236 |
+
"geometric_feature_dim": 412,
|
| 237 |
+
"kinetic_feature_dim": 90,
|
| 238 |
+
"num_samples": 102,
|
| 239 |
+
"pfc_gap": {
|
| 240 |
+
"max": 2.932789335317619,
|
| 241 |
+
"mean": 0.15156486305032094,
|
| 242 |
+
"min": -4.195418178757522,
|
| 243 |
+
"std": 0.7163983930524794
|
| 244 |
+
},
|
| 245 |
+
"pfc_generated": {
|
| 246 |
+
"max": 3.890725147350027,
|
| 247 |
+
"mean": 0.6625580218185563,
|
| 248 |
+
"min": 0.04444435017767617,
|
| 249 |
+
"std": 0.61980077587531
|
| 250 |
+
},
|
| 251 |
+
"pfc_reference": {
|
| 252 |
+
"max": 4.452664554185064,
|
| 253 |
+
"mean": 0.5109931587682353,
|
| 254 |
+
"min": 0.030597065341000908,
|
| 255 |
+
"std": 0.5892267025941194
|
| 256 |
+
}
|
| 257 |
+
},
|
| 258 |
+
"audio": {
|
| 259 |
+
"beat_align_gap": {
|
| 260 |
+
"max": 0.20317142431931395,
|
| 261 |
+
"mean": -0.0836915032734787,
|
| 262 |
+
"min": -0.40539816524995836,
|
| 263 |
+
"std": 0.09985131455347032
|
| 264 |
+
},
|
| 265 |
+
"beat_align_generated": {
|
| 266 |
+
"max": 0.6732069018318989,
|
| 267 |
+
"mean": 0.5035843145650215,
|
| 268 |
+
"min": 0.25486845614143144,
|
| 269 |
+
"std": 0.07740773730765173
|
| 270 |
+
},
|
| 271 |
+
"beat_align_reference": {
|
| 272 |
+
"max": 0.6759029031940639,
|
| 273 |
+
"mean": 0.5872758178385001,
|
| 274 |
+
"min": 0.44007268746408307,
|
| 275 |
+
"std": 0.06101503222114459
|
| 276 |
+
},
|
| 277 |
+
"num_samples": 102
|
| 278 |
+
},
|
| 279 |
+
"num_samples": 102,
|
| 280 |
+
"physical": {
|
| 281 |
+
"body_jerk_mean": {
|
| 282 |
+
"max": 151.898193359375,
|
| 283 |
+
"mean": 93.09871916677437,
|
| 284 |
+
"min": 52.08349609375,
|
| 285 |
+
"std": 22.473365730092304
|
| 286 |
+
},
|
| 287 |
+
"contact_sliding_speed": {
|
| 288 |
+
"max": 1.037505865097046,
|
| 289 |
+
"mean": 0.49518099120434594,
|
| 290 |
+
"min": 0.1495557427406311,
|
| 291 |
+
"std": 0.18642192977168262
|
| 292 |
+
},
|
| 293 |
+
"foot_ground_error": {
|
| 294 |
+
"max": 0.05694884434342384,
|
| 295 |
+
"mean": 0.019029651656655558,
|
| 296 |
+
"min": 0.003530437359586358,
|
| 297 |
+
"std": 0.010551578513320385
|
| 298 |
+
},
|
| 299 |
+
"num_samples": 102
|
| 300 |
+
}
|
| 301 |
+
},
|
| 302 |
+
"finedance_original_audio_test": {
|
| 303 |
+
"aistpp_edge": {
|
| 304 |
+
"div_g_gap": -0.11281678280093921,
|
| 305 |
+
"div_g_generated": 4.747224781473021,
|
| 306 |
+
"div_g_reference": 4.86004156427396,
|
| 307 |
+
"div_k_gap": -5.8534723963635535,
|
| 308 |
+
"div_k_generated": 13.641405650966057,
|
| 309 |
+
"div_k_reference": 19.49487804732961,
|
| 310 |
+
"fid_g": 1.707468424788388,
|
| 311 |
+
"fid_k": 40.484632925797484,
|
| 312 |
+
"geometric_feature_dim": 412,
|
| 313 |
+
"kinetic_feature_dim": 90,
|
| 314 |
+
"num_samples": 102,
|
| 315 |
+
"pfc_gap": {
|
| 316 |
+
"max": 3.6477591193684584,
|
| 317 |
+
"mean": 0.4045132325959266,
|
| 318 |
+
"min": -2.0018882941225815,
|
| 319 |
+
"std": 0.7903098367331967
|
| 320 |
+
},
|
| 321 |
+
"pfc_generated": {
|
| 322 |
+
"max": 4.414909221516321,
|
| 323 |
+
"mean": 0.7208673930521659,
|
| 324 |
+
"min": 0.01685753904491638,
|
| 325 |
+
"std": 0.8110987203020705
|
| 326 |
+
},
|
| 327 |
+
"pfc_reference": {
|
| 328 |
+
"max": 3.122786190063731,
|
| 329 |
+
"mean": 0.3163541604562393,
|
| 330 |
+
"min": 0.0007167477980582575,
|
| 331 |
+
"std": 0.4615228956341427
|
| 332 |
+
}
|
| 333 |
+
},
|
| 334 |
+
"audio": {
|
| 335 |
+
"beat_align_gap": {
|
| 336 |
+
"max": 0.5120083942239315,
|
| 337 |
+
"mean": -0.0469387823901341,
|
| 338 |
+
"min": -0.5655078705940795,
|
| 339 |
+
"std": 0.2377869692757275
|
| 340 |
+
},
|
| 341 |
+
"beat_align_generated": {
|
| 342 |
+
"max": 0.8733484359117868,
|
| 343 |
+
"mean": 0.48259422689922377,
|
| 344 |
+
"min": 0.08184839848172423,
|
| 345 |
+
"std": 0.19826141154513752
|
| 346 |
+
},
|
| 347 |
+
"beat_align_reference": {
|
| 348 |
+
"max": 0.9155656239411911,
|
| 349 |
+
"mean": 0.529533009289358,
|
| 350 |
+
"min": 0.16565130039393888,
|
| 351 |
+
"std": 0.18269913727648307
|
| 352 |
+
},
|
| 353 |
+
"num_samples": 102
|
| 354 |
+
},
|
| 355 |
+
"num_samples": 102,
|
| 356 |
+
"physical": {
|
| 357 |
+
"body_jerk_mean": {
|
| 358 |
+
"max": 208.11827087402344,
|
| 359 |
+
"mean": 101.01181910084743,
|
| 360 |
+
"min": 15.315655708312988,
|
| 361 |
+
"std": 42.45778704969256
|
| 362 |
+
},
|
| 363 |
+
"contact_sliding_speed": {
|
| 364 |
+
"max": 1.3698070049285889,
|
| 365 |
+
"mean": 0.4525775205636141,
|
| 366 |
+
"min": 0.05821777507662773,
|
| 367 |
+
"std": 0.23398819767804652
|
| 368 |
+
},
|
| 369 |
+
"foot_ground_error": {
|
| 370 |
+
"max": 0.14215940237045288,
|
| 371 |
+
"mean": 0.03325672116011893,
|
| 372 |
+
"min": 0.0028612713795155287,
|
| 373 |
+
"std": 0.029781991560662673
|
| 374 |
+
},
|
| 375 |
+
"num_samples": 102
|
| 376 |
+
}
|
| 377 |
+
},
|
| 378 |
+
"opendance_original_audio_test": {
|
| 379 |
+
"aistpp_edge": {
|
| 380 |
+
"div_g_gap": -0.03468968675710382,
|
| 381 |
+
"div_g_generated": 4.486163913297346,
|
| 382 |
+
"div_g_reference": 4.5208536000544495,
|
| 383 |
+
"div_k_gap": -2.374843698173482,
|
| 384 |
+
"div_k_generated": 8.840375829535647,
|
| 385 |
+
"div_k_reference": 11.215219527709129,
|
| 386 |
+
"fid_g": 1.590908401644036,
|
| 387 |
+
"fid_k": 18.76958645090764,
|
| 388 |
+
"geometric_feature_dim": 412,
|
| 389 |
+
"kinetic_feature_dim": 90,
|
| 390 |
+
"num_samples": 102,
|
| 391 |
+
"pfc_gap": {
|
| 392 |
+
"max": 2.2686089657121986,
|
| 393 |
+
"mean": 0.1382437120176589,
|
| 394 |
+
"min": -0.24399738769559945,
|
| 395 |
+
"std": 0.3123090105240401
|
| 396 |
+
},
|
| 397 |
+
"pfc_generated": {
|
| 398 |
+
"max": 2.3862508697288667,
|
| 399 |
+
"mean": 0.21906767735873303,
|
| 400 |
+
"min": 0.011352475937714133,
|
| 401 |
+
"std": 0.31867181439113135
|
| 402 |
+
},
|
| 403 |
+
"pfc_reference": {
|
| 404 |
+
"max": 0.5202682826002994,
|
| 405 |
+
"mean": 0.08082396534107408,
|
| 406 |
+
"min": 1.734456963144955e-05,
|
| 407 |
+
"std": 0.09160782744166214
|
| 408 |
+
}
|
| 409 |
+
},
|
| 410 |
+
"audio": {
|
| 411 |
+
"beat_align_gap": {
|
| 412 |
+
"max": 0.21636652206933799,
|
| 413 |
+
"mean": -0.09326855318539262,
|
| 414 |
+
"min": -0.4055321230244736,
|
| 415 |
+
"std": 0.10144398731338053
|
| 416 |
+
},
|
| 417 |
+
"beat_align_generated": {
|
| 418 |
+
"max": 0.691326904650172,
|
| 419 |
+
"mean": 0.4845687342862411,
|
| 420 |
+
"min": 0.0,
|
| 421 |
+
"std": 0.09971205068090859
|
| 422 |
+
},
|
| 423 |
+
"beat_align_reference": {
|
| 424 |
+
"max": 0.7346093465204763,
|
| 425 |
+
"mean": 0.5778372874716339,
|
| 426 |
+
"min": 0.0,
|
| 427 |
+
"std": 0.08261153009190644
|
| 428 |
+
},
|
| 429 |
+
"num_samples": 102
|
| 430 |
+
},
|
| 431 |
+
"num_samples": 102,
|
| 432 |
+
"physical": {
|
| 433 |
+
"body_jerk_mean": {
|
| 434 |
+
"max": 168.04806518554688,
|
| 435 |
+
"mean": 72.5217565835691,
|
| 436 |
+
"min": 18.360395431518555,
|
| 437 |
+
"std": 24.450591057024837
|
| 438 |
+
},
|
| 439 |
+
"contact_sliding_speed": {
|
| 440 |
+
"max": 0.7384104132652283,
|
| 441 |
+
"mean": 0.24727879366015687,
|
| 442 |
+
"min": 0.0463300384581089,
|
| 443 |
+
"std": 0.12928315861184075
|
| 444 |
+
},
|
| 445 |
+
"foot_ground_error": {
|
| 446 |
+
"max": 0.09883774071931839,
|
| 447 |
+
"mean": 0.07268030654785096,
|
| 448 |
+
"min": 0.019137809053063393,
|
| 449 |
+
"std": 0.013635613677582127
|
| 450 |
+
},
|
| 451 |
+
"num_samples": 102
|
| 452 |
+
}
|
| 453 |
+
}
|
| 454 |
+
},
|
| 455 |
+
"exp": "g1_transformer_guided_300m",
|
| 456 |
+
"generated_qpos": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/generated_qpos.npz",
|
| 457 |
+
"label": "OmniMotion 300M mixed-modal s055000",
|
| 458 |
+
"num_frames": 60,
|
| 459 |
+
"num_samples": 512,
|
| 460 |
+
"physical_metrics": {
|
| 461 |
+
"body_jerk_mean": {
|
| 462 |
+
"max": 278.0105285644531,
|
| 463 |
+
"mean": 95.73342476598918,
|
| 464 |
+
"min": 15.315655708312988,
|
| 465 |
+
"std": 40.634993922345
|
| 466 |
+
},
|
| 467 |
+
"contact_height_threshold": 0.12,
|
| 468 |
+
"contact_penetration_tolerance": 0.02,
|
| 469 |
+
"contact_sliding_speed": {
|
| 470 |
+
"max": 1.3698070049285889,
|
| 471 |
+
"mean": 0.40614730156812584,
|
| 472 |
+
"min": 0.0,
|
| 473 |
+
"std": 0.2216091804538725
|
| 474 |
+
},
|
| 475 |
+
"foot_ground_error": {
|
| 476 |
+
"max": 0.18515852093696594,
|
| 477 |
+
"mean": 0.040984836867210106,
|
| 478 |
+
"min": 0.0028612713795155287,
|
| 479 |
+
"std": 0.02837769267389404
|
| 480 |
+
},
|
| 481 |
+
"frames": 60,
|
| 482 |
+
"num_samples": 512
|
| 483 |
+
},
|
| 484 |
+
"reference_qpos": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/reference_qpos.npz",
|
| 485 |
+
"sample_metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/sample_metrics.jsonl",
|
| 486 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/audio_test_512.jsonl",
|
| 487 |
+
"split": "test",
|
| 488 |
+
"tracker_executed": {
|
| 489 |
+
"action_clip": 10.0,
|
| 490 |
+
"artifact_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_executed_qpos.npz",
|
| 491 |
+
"benchmark_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/benchmark_tracker.json",
|
| 492 |
+
"control_substeps": 10,
|
| 493 |
+
"enabled": true,
|
| 494 |
+
"holomotion_onnx": "${OMG_MODELS_ROOT}/HoloMotion/HoloMotion_v1.2/holomotion_v1.2_motion_tracking_model/exported/model.onnx",
|
| 495 |
+
"metrics": {
|
| 496 |
+
"e_acc": {
|
| 497 |
+
"max": 5.082680843770504,
|
| 498 |
+
"mean": 1.897563667569102,
|
| 499 |
+
"min": 0.35615143133327365,
|
| 500 |
+
"std": 0.8002793205380141
|
| 501 |
+
},
|
| 502 |
+
"e_vel": {
|
| 503 |
+
"max": 17.04840175807476,
|
| 504 |
+
"mean": 6.103366651132092,
|
| 505 |
+
"min": 1.4394153840839863,
|
| 506 |
+
"std": 2.3867622499218415
|
| 507 |
+
},
|
| 508 |
+
"enabled": true,
|
| 509 |
+
"g_mpjpe": {
|
| 510 |
+
"max": 341.753214597702,
|
| 511 |
+
"mean": 95.1012089572032,
|
| 512 |
+
"min": 28.030861169099808,
|
| 513 |
+
"std": 39.98059037197871
|
| 514 |
+
},
|
| 515 |
+
"mpjpe": {
|
| 516 |
+
"max": 144.38872039318085,
|
| 517 |
+
"mean": 41.06933371076593,
|
| 518 |
+
"min": 18.016833811998367,
|
| 519 |
+
"std": 15.862212133962863
|
| 520 |
+
},
|
| 521 |
+
"num_samples": 512,
|
| 522 |
+
"root_index": 0,
|
| 523 |
+
"target_fps": 50.0
|
| 524 |
+
},
|
| 525 |
+
"metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_metrics.json",
|
| 526 |
+
"num_generated_samples": 512,
|
| 527 |
+
"num_samples": 512,
|
| 528 |
+
"providers": "TensorrtExecutionProvider,CUDAExecutionProvider",
|
| 529 |
+
"robot_xml": null,
|
| 530 |
+
"root_index": 0,
|
| 531 |
+
"sample_metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_sample_metrics.jsonl",
|
| 532 |
+
"selected_indices": [
|
| 533 |
+
0,
|
| 534 |
+
1,
|
| 535 |
+
2,
|
| 536 |
+
3,
|
| 537 |
+
4,
|
| 538 |
+
5,
|
| 539 |
+
6,
|
| 540 |
+
7,
|
| 541 |
+
8,
|
| 542 |
+
9,
|
| 543 |
+
10,
|
| 544 |
+
11,
|
| 545 |
+
12,
|
| 546 |
+
13,
|
| 547 |
+
14,
|
| 548 |
+
15,
|
| 549 |
+
16,
|
| 550 |
+
17,
|
| 551 |
+
18,
|
| 552 |
+
19,
|
| 553 |
+
20,
|
| 554 |
+
21,
|
| 555 |
+
22,
|
| 556 |
+
23,
|
| 557 |
+
24,
|
| 558 |
+
25,
|
| 559 |
+
26,
|
| 560 |
+
27,
|
| 561 |
+
28,
|
| 562 |
+
29,
|
| 563 |
+
30,
|
| 564 |
+
31,
|
| 565 |
+
32,
|
| 566 |
+
33,
|
| 567 |
+
34,
|
| 568 |
+
35,
|
| 569 |
+
36,
|
| 570 |
+
37,
|
| 571 |
+
38,
|
| 572 |
+
39,
|
| 573 |
+
40,
|
| 574 |
+
41,
|
| 575 |
+
42,
|
| 576 |
+
43,
|
| 577 |
+
44,
|
| 578 |
+
45,
|
| 579 |
+
46,
|
| 580 |
+
47,
|
| 581 |
+
48,
|
| 582 |
+
49,
|
| 583 |
+
50,
|
| 584 |
+
51,
|
| 585 |
+
52,
|
| 586 |
+
53,
|
| 587 |
+
54,
|
| 588 |
+
55,
|
| 589 |
+
56,
|
| 590 |
+
57,
|
| 591 |
+
58,
|
| 592 |
+
59,
|
| 593 |
+
60,
|
| 594 |
+
61,
|
| 595 |
+
62,
|
| 596 |
+
63,
|
| 597 |
+
64,
|
| 598 |
+
65,
|
| 599 |
+
66,
|
| 600 |
+
67,
|
| 601 |
+
68,
|
| 602 |
+
69,
|
| 603 |
+
70,
|
| 604 |
+
71,
|
| 605 |
+
72,
|
| 606 |
+
73,
|
| 607 |
+
74,
|
| 608 |
+
75,
|
| 609 |
+
76,
|
| 610 |
+
77,
|
| 611 |
+
78,
|
| 612 |
+
79,
|
| 613 |
+
80,
|
| 614 |
+
81,
|
| 615 |
+
82,
|
| 616 |
+
83,
|
| 617 |
+
84,
|
| 618 |
+
85,
|
| 619 |
+
86,
|
| 620 |
+
87,
|
| 621 |
+
88,
|
| 622 |
+
89,
|
| 623 |
+
90,
|
| 624 |
+
91,
|
| 625 |
+
92,
|
| 626 |
+
93,
|
| 627 |
+
94,
|
| 628 |
+
95,
|
| 629 |
+
96,
|
| 630 |
+
97,
|
| 631 |
+
98,
|
| 632 |
+
99,
|
| 633 |
+
100,
|
| 634 |
+
101,
|
| 635 |
+
102,
|
| 636 |
+
103,
|
| 637 |
+
104,
|
| 638 |
+
105,
|
| 639 |
+
106,
|
| 640 |
+
107,
|
| 641 |
+
108,
|
| 642 |
+
109,
|
| 643 |
+
110,
|
| 644 |
+
111,
|
| 645 |
+
112,
|
| 646 |
+
113,
|
| 647 |
+
114,
|
| 648 |
+
115,
|
| 649 |
+
116,
|
| 650 |
+
117,
|
| 651 |
+
118,
|
| 652 |
+
119,
|
| 653 |
+
120,
|
| 654 |
+
121,
|
| 655 |
+
122,
|
| 656 |
+
123,
|
| 657 |
+
124,
|
| 658 |
+
125,
|
| 659 |
+
126,
|
| 660 |
+
127,
|
| 661 |
+
128,
|
| 662 |
+
129,
|
| 663 |
+
130,
|
| 664 |
+
131,
|
| 665 |
+
132,
|
| 666 |
+
133,
|
| 667 |
+
134,
|
| 668 |
+
135,
|
| 669 |
+
136,
|
| 670 |
+
137,
|
| 671 |
+
138,
|
| 672 |
+
139,
|
| 673 |
+
140,
|
| 674 |
+
141,
|
| 675 |
+
142,
|
| 676 |
+
143,
|
| 677 |
+
144,
|
| 678 |
+
145,
|
| 679 |
+
146,
|
| 680 |
+
147,
|
| 681 |
+
148,
|
| 682 |
+
149,
|
| 683 |
+
150,
|
| 684 |
+
151,
|
| 685 |
+
152,
|
| 686 |
+
153,
|
| 687 |
+
154,
|
| 688 |
+
155,
|
| 689 |
+
156,
|
| 690 |
+
157,
|
| 691 |
+
158,
|
| 692 |
+
159,
|
| 693 |
+
160,
|
| 694 |
+
161,
|
| 695 |
+
162,
|
| 696 |
+
163,
|
| 697 |
+
164,
|
| 698 |
+
165,
|
| 699 |
+
166,
|
| 700 |
+
167,
|
| 701 |
+
168,
|
| 702 |
+
169,
|
| 703 |
+
170,
|
| 704 |
+
171,
|
| 705 |
+
172,
|
| 706 |
+
173,
|
| 707 |
+
174,
|
| 708 |
+
175,
|
| 709 |
+
176,
|
| 710 |
+
177,
|
| 711 |
+
178,
|
| 712 |
+
179,
|
| 713 |
+
180,
|
| 714 |
+
181,
|
| 715 |
+
182,
|
| 716 |
+
183,
|
| 717 |
+
184,
|
| 718 |
+
185,
|
| 719 |
+
186,
|
| 720 |
+
187,
|
| 721 |
+
188,
|
| 722 |
+
189,
|
| 723 |
+
190,
|
| 724 |
+
191,
|
| 725 |
+
192,
|
| 726 |
+
193,
|
| 727 |
+
194,
|
| 728 |
+
195,
|
| 729 |
+
196,
|
| 730 |
+
197,
|
| 731 |
+
198,
|
| 732 |
+
199,
|
| 733 |
+
200,
|
| 734 |
+
201,
|
| 735 |
+
202,
|
| 736 |
+
203,
|
| 737 |
+
204,
|
| 738 |
+
205,
|
| 739 |
+
206,
|
| 740 |
+
207,
|
| 741 |
+
208,
|
| 742 |
+
209,
|
| 743 |
+
210,
|
| 744 |
+
211,
|
| 745 |
+
212,
|
| 746 |
+
213,
|
| 747 |
+
214,
|
| 748 |
+
215,
|
| 749 |
+
216,
|
| 750 |
+
217,
|
| 751 |
+
218,
|
| 752 |
+
219,
|
| 753 |
+
220,
|
| 754 |
+
221,
|
| 755 |
+
222,
|
| 756 |
+
223,
|
| 757 |
+
224,
|
| 758 |
+
225,
|
| 759 |
+
226,
|
| 760 |
+
227,
|
| 761 |
+
228,
|
| 762 |
+
229,
|
| 763 |
+
230,
|
| 764 |
+
231,
|
| 765 |
+
232,
|
| 766 |
+
233,
|
| 767 |
+
234,
|
| 768 |
+
235,
|
| 769 |
+
236,
|
| 770 |
+
237,
|
| 771 |
+
238,
|
| 772 |
+
239,
|
| 773 |
+
240,
|
| 774 |
+
241,
|
| 775 |
+
242,
|
| 776 |
+
243,
|
| 777 |
+
244,
|
| 778 |
+
245,
|
| 779 |
+
246,
|
| 780 |
+
247,
|
| 781 |
+
248,
|
| 782 |
+
249,
|
| 783 |
+
250,
|
| 784 |
+
251,
|
| 785 |
+
252,
|
| 786 |
+
253,
|
| 787 |
+
254,
|
| 788 |
+
255,
|
| 789 |
+
256,
|
| 790 |
+
257,
|
| 791 |
+
258,
|
| 792 |
+
259,
|
| 793 |
+
260,
|
| 794 |
+
261,
|
| 795 |
+
262,
|
| 796 |
+
263,
|
| 797 |
+
264,
|
| 798 |
+
265,
|
| 799 |
+
266,
|
| 800 |
+
267,
|
| 801 |
+
268,
|
| 802 |
+
269,
|
| 803 |
+
270,
|
| 804 |
+
271,
|
| 805 |
+
272,
|
| 806 |
+
273,
|
| 807 |
+
274,
|
| 808 |
+
275,
|
| 809 |
+
276,
|
| 810 |
+
277,
|
| 811 |
+
278,
|
| 812 |
+
279,
|
| 813 |
+
280,
|
| 814 |
+
281,
|
| 815 |
+
282,
|
| 816 |
+
283,
|
| 817 |
+
284,
|
| 818 |
+
285,
|
| 819 |
+
286,
|
| 820 |
+
287,
|
| 821 |
+
288,
|
| 822 |
+
289,
|
| 823 |
+
290,
|
| 824 |
+
291,
|
| 825 |
+
292,
|
| 826 |
+
293,
|
| 827 |
+
294,
|
| 828 |
+
295,
|
| 829 |
+
296,
|
| 830 |
+
297,
|
| 831 |
+
298,
|
| 832 |
+
299,
|
| 833 |
+
300,
|
| 834 |
+
301,
|
| 835 |
+
302,
|
| 836 |
+
303,
|
| 837 |
+
304,
|
| 838 |
+
305,
|
| 839 |
+
306,
|
| 840 |
+
307,
|
| 841 |
+
308,
|
| 842 |
+
309,
|
| 843 |
+
310,
|
| 844 |
+
311,
|
| 845 |
+
312,
|
| 846 |
+
313,
|
| 847 |
+
314,
|
| 848 |
+
315,
|
| 849 |
+
316,
|
| 850 |
+
317,
|
| 851 |
+
318,
|
| 852 |
+
319,
|
| 853 |
+
320,
|
| 854 |
+
321,
|
| 855 |
+
322,
|
| 856 |
+
323,
|
| 857 |
+
324,
|
| 858 |
+
325,
|
| 859 |
+
326,
|
| 860 |
+
327,
|
| 861 |
+
328,
|
| 862 |
+
329,
|
| 863 |
+
330,
|
| 864 |
+
331,
|
| 865 |
+
332,
|
| 866 |
+
333,
|
| 867 |
+
334,
|
| 868 |
+
335,
|
| 869 |
+
336,
|
| 870 |
+
337,
|
| 871 |
+
338,
|
| 872 |
+
339,
|
| 873 |
+
340,
|
| 874 |
+
341,
|
| 875 |
+
342,
|
| 876 |
+
343,
|
| 877 |
+
344,
|
| 878 |
+
345,
|
| 879 |
+
346,
|
| 880 |
+
347,
|
| 881 |
+
348,
|
| 882 |
+
349,
|
| 883 |
+
350,
|
| 884 |
+
351,
|
| 885 |
+
352,
|
| 886 |
+
353,
|
| 887 |
+
354,
|
| 888 |
+
355,
|
| 889 |
+
356,
|
| 890 |
+
357,
|
| 891 |
+
358,
|
| 892 |
+
359,
|
| 893 |
+
360,
|
| 894 |
+
361,
|
| 895 |
+
362,
|
| 896 |
+
363,
|
| 897 |
+
364,
|
| 898 |
+
365,
|
| 899 |
+
366,
|
| 900 |
+
367,
|
| 901 |
+
368,
|
| 902 |
+
369,
|
| 903 |
+
370,
|
| 904 |
+
371,
|
| 905 |
+
372,
|
| 906 |
+
373,
|
| 907 |
+
374,
|
| 908 |
+
375,
|
| 909 |
+
376,
|
| 910 |
+
377,
|
| 911 |
+
378,
|
| 912 |
+
379,
|
| 913 |
+
380,
|
| 914 |
+
381,
|
| 915 |
+
382,
|
| 916 |
+
383,
|
| 917 |
+
384,
|
| 918 |
+
385,
|
| 919 |
+
386,
|
| 920 |
+
387,
|
| 921 |
+
388,
|
| 922 |
+
389,
|
| 923 |
+
390,
|
| 924 |
+
391,
|
| 925 |
+
392,
|
| 926 |
+
393,
|
| 927 |
+
394,
|
| 928 |
+
395,
|
| 929 |
+
396,
|
| 930 |
+
397,
|
| 931 |
+
398,
|
| 932 |
+
399,
|
| 933 |
+
400,
|
| 934 |
+
401,
|
| 935 |
+
402,
|
| 936 |
+
403,
|
| 937 |
+
404,
|
| 938 |
+
405,
|
| 939 |
+
406,
|
| 940 |
+
407,
|
| 941 |
+
408,
|
| 942 |
+
409,
|
| 943 |
+
410,
|
| 944 |
+
411,
|
| 945 |
+
412,
|
| 946 |
+
413,
|
| 947 |
+
414,
|
| 948 |
+
415,
|
| 949 |
+
416,
|
| 950 |
+
417,
|
| 951 |
+
418,
|
| 952 |
+
419,
|
| 953 |
+
420,
|
| 954 |
+
421,
|
| 955 |
+
422,
|
| 956 |
+
423,
|
| 957 |
+
424,
|
| 958 |
+
425,
|
| 959 |
+
426,
|
| 960 |
+
427,
|
| 961 |
+
428,
|
| 962 |
+
429,
|
| 963 |
+
430,
|
| 964 |
+
431,
|
| 965 |
+
432,
|
| 966 |
+
433,
|
| 967 |
+
434,
|
| 968 |
+
435,
|
| 969 |
+
436,
|
| 970 |
+
437,
|
| 971 |
+
438,
|
| 972 |
+
439,
|
| 973 |
+
440,
|
| 974 |
+
441,
|
| 975 |
+
442,
|
| 976 |
+
443,
|
| 977 |
+
444,
|
| 978 |
+
445,
|
| 979 |
+
446,
|
| 980 |
+
447,
|
| 981 |
+
448,
|
| 982 |
+
449,
|
| 983 |
+
450,
|
| 984 |
+
451,
|
| 985 |
+
452,
|
| 986 |
+
453,
|
| 987 |
+
454,
|
| 988 |
+
455,
|
| 989 |
+
456,
|
| 990 |
+
457,
|
| 991 |
+
458,
|
| 992 |
+
459,
|
| 993 |
+
460,
|
| 994 |
+
461,
|
| 995 |
+
462,
|
| 996 |
+
463,
|
| 997 |
+
464,
|
| 998 |
+
465,
|
| 999 |
+
466,
|
| 1000 |
+
467,
|
| 1001 |
+
468,
|
| 1002 |
+
469,
|
| 1003 |
+
470,
|
| 1004 |
+
471,
|
| 1005 |
+
472,
|
| 1006 |
+
473,
|
| 1007 |
+
474,
|
| 1008 |
+
475,
|
| 1009 |
+
476,
|
| 1010 |
+
477,
|
| 1011 |
+
478,
|
| 1012 |
+
479,
|
| 1013 |
+
480,
|
| 1014 |
+
481,
|
| 1015 |
+
482,
|
| 1016 |
+
483,
|
| 1017 |
+
484,
|
| 1018 |
+
485,
|
| 1019 |
+
486,
|
| 1020 |
+
487,
|
| 1021 |
+
488,
|
| 1022 |
+
489,
|
| 1023 |
+
490,
|
| 1024 |
+
491,
|
| 1025 |
+
492,
|
| 1026 |
+
493,
|
| 1027 |
+
494,
|
| 1028 |
+
495,
|
| 1029 |
+
496,
|
| 1030 |
+
497,
|
| 1031 |
+
498,
|
| 1032 |
+
499,
|
| 1033 |
+
500,
|
| 1034 |
+
501,
|
| 1035 |
+
502,
|
| 1036 |
+
503,
|
| 1037 |
+
504,
|
| 1038 |
+
505,
|
| 1039 |
+
506,
|
| 1040 |
+
507,
|
| 1041 |
+
508,
|
| 1042 |
+
509,
|
| 1043 |
+
510,
|
| 1044 |
+
511
|
| 1045 |
+
],
|
| 1046 |
+
"steps": null,
|
| 1047 |
+
"target_fps": 50.0,
|
| 1048 |
+
"valid_shape": [
|
| 1049 |
+
512,
|
| 1050 |
+
99
|
| 1051 |
+
]
|
| 1052 |
+
},
|
| 1053 |
+
"transition_metrics": {
|
| 1054 |
+
"chunk_length": 60,
|
| 1055 |
+
"enabled": false,
|
| 1056 |
+
"reason": "disabled"
|
| 1057 |
+
}
|
| 1058 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/dataset_metrics.json
ADDED
|
@@ -0,0 +1,382 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"aioz_gdance_original_audio_test": {
|
| 3 |
+
"aistpp_edge": {
|
| 4 |
+
"div_g_gap": 0.23862403086776185,
|
| 5 |
+
"div_g_generated": 3.7473771961157687,
|
| 6 |
+
"div_g_reference": 3.508753165248007,
|
| 7 |
+
"div_k_gap": -1.4359521145663674,
|
| 8 |
+
"div_k_generated": 9.315809033822825,
|
| 9 |
+
"div_k_reference": 10.751761148389193,
|
| 10 |
+
"fid_g": 1.4318985574865248,
|
| 11 |
+
"fid_k": 7.601990475458294,
|
| 12 |
+
"geometric_feature_dim": 412,
|
| 13 |
+
"kinetic_feature_dim": 90,
|
| 14 |
+
"num_samples": 103,
|
| 15 |
+
"pfc_gap": {
|
| 16 |
+
"max": 3.2058528208797874,
|
| 17 |
+
"mean": 0.05262750505804416,
|
| 18 |
+
"min": -1.6349628441701682,
|
| 19 |
+
"std": 0.6426666823502454
|
| 20 |
+
},
|
| 21 |
+
"pfc_generated": {
|
| 22 |
+
"max": 3.4774407560932494,
|
| 23 |
+
"mean": 0.3836619604251987,
|
| 24 |
+
"min": 0.005071079598004716,
|
| 25 |
+
"std": 0.5183719007045057
|
| 26 |
+
},
|
| 27 |
+
"pfc_reference": {
|
| 28 |
+
"max": 1.8273494537376207,
|
| 29 |
+
"mean": 0.3310344553671544,
|
| 30 |
+
"min": 0.0002850814965562205,
|
| 31 |
+
"std": 0.42521448548204027
|
| 32 |
+
}
|
| 33 |
+
},
|
| 34 |
+
"audio": {
|
| 35 |
+
"beat_align_gap": {
|
| 36 |
+
"max": 0.23840448764256933,
|
| 37 |
+
"mean": -0.033628672950055886,
|
| 38 |
+
"min": -0.32453708581516766,
|
| 39 |
+
"std": 0.13742288780504877
|
| 40 |
+
},
|
| 41 |
+
"beat_align_generated": {
|
| 42 |
+
"max": 0.6725372901371526,
|
| 43 |
+
"mean": 0.4919088098323277,
|
| 44 |
+
"min": 0.24963270016294453,
|
| 45 |
+
"std": 0.09099816858378701
|
| 46 |
+
},
|
| 47 |
+
"beat_align_reference": {
|
| 48 |
+
"max": 0.723284948504687,
|
| 49 |
+
"mean": 0.5255374827823837,
|
| 50 |
+
"min": 0.33650493752041727,
|
| 51 |
+
"std": 0.08282468265444852
|
| 52 |
+
},
|
| 53 |
+
"num_samples": 103
|
| 54 |
+
},
|
| 55 |
+
"num_samples": 103,
|
| 56 |
+
"physical": {
|
| 57 |
+
"body_jerk_mean": {
|
| 58 |
+
"max": 187.8517608642578,
|
| 59 |
+
"mean": 75.8366039979805,
|
| 60 |
+
"min": 27.45358657836914,
|
| 61 |
+
"std": 24.36331538169348
|
| 62 |
+
},
|
| 63 |
+
"contact_sliding_speed": {
|
| 64 |
+
"max": 0.8738095164299011,
|
| 65 |
+
"mean": 0.3195253503915754,
|
| 66 |
+
"min": 0.043544918298721313,
|
| 67 |
+
"std": 0.16198516039471123
|
| 68 |
+
},
|
| 69 |
+
"foot_ground_error": {
|
| 70 |
+
"max": 0.08582238852977753,
|
| 71 |
+
"mean": 0.04983966779054224,
|
| 72 |
+
"min": 0.007044824305921793,
|
| 73 |
+
"std": 0.017409112859072134
|
| 74 |
+
},
|
| 75 |
+
"num_samples": 103
|
| 76 |
+
}
|
| 77 |
+
},
|
| 78 |
+
"aistpp_original_audio_test": {
|
| 79 |
+
"aistpp_edge": {
|
| 80 |
+
"div_g_gap": 0.3749239056934046,
|
| 81 |
+
"div_g_generated": 3.929406542818343,
|
| 82 |
+
"div_g_reference": 3.5544826371249383,
|
| 83 |
+
"div_k_gap": -12.850860561893567,
|
| 84 |
+
"div_k_generated": 12.909821306156998,
|
| 85 |
+
"div_k_reference": 25.760681868050565,
|
| 86 |
+
"fid_g": 2.601341460428607,
|
| 87 |
+
"fid_k": 251.18791168722862,
|
| 88 |
+
"geometric_feature_dim": 412,
|
| 89 |
+
"kinetic_feature_dim": 90,
|
| 90 |
+
"num_samples": 103,
|
| 91 |
+
"pfc_gap": {
|
| 92 |
+
"max": 2.477920920125234,
|
| 93 |
+
"mean": -0.24784057237073714,
|
| 94 |
+
"min": -4.8420685154923095,
|
| 95 |
+
"std": 1.3450989582965898
|
| 96 |
+
},
|
| 97 |
+
"pfc_generated": {
|
| 98 |
+
"max": 3.0173064117838173,
|
| 99 |
+
"mean": 0.8859167056460855,
|
| 100 |
+
"min": 0.03606813460559294,
|
| 101 |
+
"std": 0.6825868113283983
|
| 102 |
+
},
|
| 103 |
+
"pfc_reference": {
|
| 104 |
+
"max": 5.707070070036645,
|
| 105 |
+
"mean": 1.1337572780168226,
|
| 106 |
+
"min": 7.906044911863323e-05,
|
| 107 |
+
"std": 1.2009778419653923
|
| 108 |
+
}
|
| 109 |
+
},
|
| 110 |
+
"audio": {
|
| 111 |
+
"beat_align_gap": {
|
| 112 |
+
"max": 0.5701364518161618,
|
| 113 |
+
"mean": -0.0801368416977987,
|
| 114 |
+
"min": -0.7608427161001056,
|
| 115 |
+
"std": 0.28185526909472197
|
| 116 |
+
},
|
| 117 |
+
"beat_align_generated": {
|
| 118 |
+
"max": 0.9819864896355885,
|
| 119 |
+
"mean": 0.5061291397536287,
|
| 120 |
+
"min": 0.0,
|
| 121 |
+
"std": 0.2064175413162069
|
| 122 |
+
},
|
| 123 |
+
"beat_align_reference": {
|
| 124 |
+
"max": 0.9864898672266914,
|
| 125 |
+
"mean": 0.5862659814514274,
|
| 126 |
+
"min": 0.0,
|
| 127 |
+
"std": 0.19459278456334647
|
| 128 |
+
},
|
| 129 |
+
"num_samples": 103
|
| 130 |
+
},
|
| 131 |
+
"num_samples": 103,
|
| 132 |
+
"physical": {
|
| 133 |
+
"body_jerk_mean": {
|
| 134 |
+
"max": 278.0105285644531,
|
| 135 |
+
"mean": 135.99853585993202,
|
| 136 |
+
"min": 32.689208984375,
|
| 137 |
+
"std": 46.387850301317464
|
| 138 |
+
},
|
| 139 |
+
"contact_sliding_speed": {
|
| 140 |
+
"max": 1.3322360515594482,
|
| 141 |
+
"mean": 0.5159466229017499,
|
| 142 |
+
"min": 0.0,
|
| 143 |
+
"std": 0.24116125404927893
|
| 144 |
+
},
|
| 145 |
+
"foot_ground_error": {
|
| 146 |
+
"max": 0.18515852093696594,
|
| 147 |
+
"mean": 0.030137372799941057,
|
| 148 |
+
"min": 0.004513147287070751,
|
| 149 |
+
"std": 0.028280388956618472
|
| 150 |
+
},
|
| 151 |
+
"num_samples": 103
|
| 152 |
+
}
|
| 153 |
+
},
|
| 154 |
+
"compas3d_original_audio_test": {
|
| 155 |
+
"aistpp_edge": {
|
| 156 |
+
"div_g_gap": 0.17526199534596376,
|
| 157 |
+
"div_g_generated": 4.703408486137112,
|
| 158 |
+
"div_g_reference": 4.528146490791149,
|
| 159 |
+
"div_k_gap": -3.1528614038078047,
|
| 160 |
+
"div_k_generated": 8.431761165269869,
|
| 161 |
+
"div_k_reference": 11.584622569077673,
|
| 162 |
+
"fid_g": 2.805483949908389,
|
| 163 |
+
"fid_k": 67.40567433330911,
|
| 164 |
+
"geometric_feature_dim": 412,
|
| 165 |
+
"kinetic_feature_dim": 90,
|
| 166 |
+
"num_samples": 102,
|
| 167 |
+
"pfc_gap": {
|
| 168 |
+
"max": 2.932789335317619,
|
| 169 |
+
"mean": 0.15156486305032094,
|
| 170 |
+
"min": -4.195418178757522,
|
| 171 |
+
"std": 0.7163983930524794
|
| 172 |
+
},
|
| 173 |
+
"pfc_generated": {
|
| 174 |
+
"max": 3.890725147350027,
|
| 175 |
+
"mean": 0.6625580218185563,
|
| 176 |
+
"min": 0.04444435017767617,
|
| 177 |
+
"std": 0.61980077587531
|
| 178 |
+
},
|
| 179 |
+
"pfc_reference": {
|
| 180 |
+
"max": 4.452664554185064,
|
| 181 |
+
"mean": 0.5109931587682353,
|
| 182 |
+
"min": 0.030597065341000908,
|
| 183 |
+
"std": 0.5892267025941194
|
| 184 |
+
}
|
| 185 |
+
},
|
| 186 |
+
"audio": {
|
| 187 |
+
"beat_align_gap": {
|
| 188 |
+
"max": 0.20317142431931395,
|
| 189 |
+
"mean": -0.0836915032734787,
|
| 190 |
+
"min": -0.40539816524995836,
|
| 191 |
+
"std": 0.09985131455347032
|
| 192 |
+
},
|
| 193 |
+
"beat_align_generated": {
|
| 194 |
+
"max": 0.6732069018318989,
|
| 195 |
+
"mean": 0.5035843145650215,
|
| 196 |
+
"min": 0.25486845614143144,
|
| 197 |
+
"std": 0.07740773730765173
|
| 198 |
+
},
|
| 199 |
+
"beat_align_reference": {
|
| 200 |
+
"max": 0.6759029031940639,
|
| 201 |
+
"mean": 0.5872758178385001,
|
| 202 |
+
"min": 0.44007268746408307,
|
| 203 |
+
"std": 0.06101503222114459
|
| 204 |
+
},
|
| 205 |
+
"num_samples": 102
|
| 206 |
+
},
|
| 207 |
+
"num_samples": 102,
|
| 208 |
+
"physical": {
|
| 209 |
+
"body_jerk_mean": {
|
| 210 |
+
"max": 151.898193359375,
|
| 211 |
+
"mean": 93.09871916677437,
|
| 212 |
+
"min": 52.08349609375,
|
| 213 |
+
"std": 22.473365730092304
|
| 214 |
+
},
|
| 215 |
+
"contact_sliding_speed": {
|
| 216 |
+
"max": 1.037505865097046,
|
| 217 |
+
"mean": 0.49518099120434594,
|
| 218 |
+
"min": 0.1495557427406311,
|
| 219 |
+
"std": 0.18642192977168262
|
| 220 |
+
},
|
| 221 |
+
"foot_ground_error": {
|
| 222 |
+
"max": 0.05694884434342384,
|
| 223 |
+
"mean": 0.019029651656655558,
|
| 224 |
+
"min": 0.003530437359586358,
|
| 225 |
+
"std": 0.010551578513320385
|
| 226 |
+
},
|
| 227 |
+
"num_samples": 102
|
| 228 |
+
}
|
| 229 |
+
},
|
| 230 |
+
"finedance_original_audio_test": {
|
| 231 |
+
"aistpp_edge": {
|
| 232 |
+
"div_g_gap": -0.11281678280093921,
|
| 233 |
+
"div_g_generated": 4.747224781473021,
|
| 234 |
+
"div_g_reference": 4.86004156427396,
|
| 235 |
+
"div_k_gap": -5.8534723963635535,
|
| 236 |
+
"div_k_generated": 13.641405650966057,
|
| 237 |
+
"div_k_reference": 19.49487804732961,
|
| 238 |
+
"fid_g": 1.707468424788388,
|
| 239 |
+
"fid_k": 40.484632925797484,
|
| 240 |
+
"geometric_feature_dim": 412,
|
| 241 |
+
"kinetic_feature_dim": 90,
|
| 242 |
+
"num_samples": 102,
|
| 243 |
+
"pfc_gap": {
|
| 244 |
+
"max": 3.6477591193684584,
|
| 245 |
+
"mean": 0.4045132325959266,
|
| 246 |
+
"min": -2.0018882941225815,
|
| 247 |
+
"std": 0.7903098367331967
|
| 248 |
+
},
|
| 249 |
+
"pfc_generated": {
|
| 250 |
+
"max": 4.414909221516321,
|
| 251 |
+
"mean": 0.7208673930521659,
|
| 252 |
+
"min": 0.01685753904491638,
|
| 253 |
+
"std": 0.8110987203020705
|
| 254 |
+
},
|
| 255 |
+
"pfc_reference": {
|
| 256 |
+
"max": 3.122786190063731,
|
| 257 |
+
"mean": 0.3163541604562393,
|
| 258 |
+
"min": 0.0007167477980582575,
|
| 259 |
+
"std": 0.4615228956341427
|
| 260 |
+
}
|
| 261 |
+
},
|
| 262 |
+
"audio": {
|
| 263 |
+
"beat_align_gap": {
|
| 264 |
+
"max": 0.5120083942239315,
|
| 265 |
+
"mean": -0.0469387823901341,
|
| 266 |
+
"min": -0.5655078705940795,
|
| 267 |
+
"std": 0.2377869692757275
|
| 268 |
+
},
|
| 269 |
+
"beat_align_generated": {
|
| 270 |
+
"max": 0.8733484359117868,
|
| 271 |
+
"mean": 0.48259422689922377,
|
| 272 |
+
"min": 0.08184839848172423,
|
| 273 |
+
"std": 0.19826141154513752
|
| 274 |
+
},
|
| 275 |
+
"beat_align_reference": {
|
| 276 |
+
"max": 0.9155656239411911,
|
| 277 |
+
"mean": 0.529533009289358,
|
| 278 |
+
"min": 0.16565130039393888,
|
| 279 |
+
"std": 0.18269913727648307
|
| 280 |
+
},
|
| 281 |
+
"num_samples": 102
|
| 282 |
+
},
|
| 283 |
+
"num_samples": 102,
|
| 284 |
+
"physical": {
|
| 285 |
+
"body_jerk_mean": {
|
| 286 |
+
"max": 208.11827087402344,
|
| 287 |
+
"mean": 101.01181910084743,
|
| 288 |
+
"min": 15.315655708312988,
|
| 289 |
+
"std": 42.45778704969256
|
| 290 |
+
},
|
| 291 |
+
"contact_sliding_speed": {
|
| 292 |
+
"max": 1.3698070049285889,
|
| 293 |
+
"mean": 0.4525775205636141,
|
| 294 |
+
"min": 0.05821777507662773,
|
| 295 |
+
"std": 0.23398819767804652
|
| 296 |
+
},
|
| 297 |
+
"foot_ground_error": {
|
| 298 |
+
"max": 0.14215940237045288,
|
| 299 |
+
"mean": 0.03325672116011893,
|
| 300 |
+
"min": 0.0028612713795155287,
|
| 301 |
+
"std": 0.029781991560662673
|
| 302 |
+
},
|
| 303 |
+
"num_samples": 102
|
| 304 |
+
}
|
| 305 |
+
},
|
| 306 |
+
"opendance_original_audio_test": {
|
| 307 |
+
"aistpp_edge": {
|
| 308 |
+
"div_g_gap": -0.03468968675710382,
|
| 309 |
+
"div_g_generated": 4.486163913297346,
|
| 310 |
+
"div_g_reference": 4.5208536000544495,
|
| 311 |
+
"div_k_gap": -2.374843698173482,
|
| 312 |
+
"div_k_generated": 8.840375829535647,
|
| 313 |
+
"div_k_reference": 11.215219527709129,
|
| 314 |
+
"fid_g": 1.590908401644036,
|
| 315 |
+
"fid_k": 18.76958645090764,
|
| 316 |
+
"geometric_feature_dim": 412,
|
| 317 |
+
"kinetic_feature_dim": 90,
|
| 318 |
+
"num_samples": 102,
|
| 319 |
+
"pfc_gap": {
|
| 320 |
+
"max": 2.2686089657121986,
|
| 321 |
+
"mean": 0.1382437120176589,
|
| 322 |
+
"min": -0.24399738769559945,
|
| 323 |
+
"std": 0.3123090105240401
|
| 324 |
+
},
|
| 325 |
+
"pfc_generated": {
|
| 326 |
+
"max": 2.3862508697288667,
|
| 327 |
+
"mean": 0.21906767735873303,
|
| 328 |
+
"min": 0.011352475937714133,
|
| 329 |
+
"std": 0.31867181439113135
|
| 330 |
+
},
|
| 331 |
+
"pfc_reference": {
|
| 332 |
+
"max": 0.5202682826002994,
|
| 333 |
+
"mean": 0.08082396534107408,
|
| 334 |
+
"min": 1.734456963144955e-05,
|
| 335 |
+
"std": 0.09160782744166214
|
| 336 |
+
}
|
| 337 |
+
},
|
| 338 |
+
"audio": {
|
| 339 |
+
"beat_align_gap": {
|
| 340 |
+
"max": 0.21636652206933799,
|
| 341 |
+
"mean": -0.09326855318539262,
|
| 342 |
+
"min": -0.4055321230244736,
|
| 343 |
+
"std": 0.10144398731338053
|
| 344 |
+
},
|
| 345 |
+
"beat_align_generated": {
|
| 346 |
+
"max": 0.691326904650172,
|
| 347 |
+
"mean": 0.4845687342862411,
|
| 348 |
+
"min": 0.0,
|
| 349 |
+
"std": 0.09971205068090859
|
| 350 |
+
},
|
| 351 |
+
"beat_align_reference": {
|
| 352 |
+
"max": 0.7346093465204763,
|
| 353 |
+
"mean": 0.5778372874716339,
|
| 354 |
+
"min": 0.0,
|
| 355 |
+
"std": 0.08261153009190644
|
| 356 |
+
},
|
| 357 |
+
"num_samples": 102
|
| 358 |
+
},
|
| 359 |
+
"num_samples": 102,
|
| 360 |
+
"physical": {
|
| 361 |
+
"body_jerk_mean": {
|
| 362 |
+
"max": 168.04806518554688,
|
| 363 |
+
"mean": 72.5217565835691,
|
| 364 |
+
"min": 18.360395431518555,
|
| 365 |
+
"std": 24.450591057024837
|
| 366 |
+
},
|
| 367 |
+
"contact_sliding_speed": {
|
| 368 |
+
"max": 0.7384104132652283,
|
| 369 |
+
"mean": 0.24727879366015687,
|
| 370 |
+
"min": 0.0463300384581089,
|
| 371 |
+
"std": 0.12928315861184075
|
| 372 |
+
},
|
| 373 |
+
"foot_ground_error": {
|
| 374 |
+
"max": 0.09883774071931839,
|
| 375 |
+
"mean": 0.07268030654785096,
|
| 376 |
+
"min": 0.019137809053063393,
|
| 377 |
+
"std": 0.013635613677582127
|
| 378 |
+
},
|
| 379 |
+
"num_samples": 102
|
| 380 |
+
}
|
| 381 |
+
}
|
| 382 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/merge_summary.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/audio_test_512.jsonl",
|
| 3 |
+
"output_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/generated_qpos.npz",
|
| 4 |
+
"num_samples": 512,
|
| 5 |
+
"num_frames": 60,
|
| 6 |
+
"num_shards": 8,
|
| 7 |
+
"shard_artifacts": [
|
| 8 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_000/generated_qpos.npz",
|
| 9 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_001/generated_qpos.npz",
|
| 10 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_002/generated_qpos.npz",
|
| 11 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_003/generated_qpos.npz",
|
| 12 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_004/generated_qpos.npz",
|
| 13 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_005/generated_qpos.npz",
|
| 14 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_006/generated_qpos.npz",
|
| 15 |
+
"${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_007/generated_qpos.npz"
|
| 16 |
+
]
|
| 17 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/metrics.json
ADDED
|
@@ -0,0 +1,1061 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark": {
|
| 3 |
+
"aistpp_edge_metrics": {
|
| 4 |
+
"div_g_gap": -0.13676034236105217,
|
| 5 |
+
"div_g_generated": 4.866377286829666,
|
| 6 |
+
"div_g_reference": 5.003137629190718,
|
| 7 |
+
"div_k_gap": -6.90296188275374,
|
| 8 |
+
"div_k_generated": 10.652930520737065,
|
| 9 |
+
"div_k_reference": 17.555892403490805,
|
| 10 |
+
"fid_g": 0.7416922948407674,
|
| 11 |
+
"fid_k": 59.19689074372258,
|
| 12 |
+
"frames": 60,
|
| 13 |
+
"geometric_feature": "g1_root_relative_body_geometry",
|
| 14 |
+
"geometric_feature_dim": 412,
|
| 15 |
+
"kinetic_feature": "fairmotion_root_relative_joint_kinetic_energy",
|
| 16 |
+
"kinetic_feature_dim": 90,
|
| 17 |
+
"left_foot_indices": [
|
| 18 |
+
5,
|
| 19 |
+
6
|
| 20 |
+
],
|
| 21 |
+
"num_samples": 512,
|
| 22 |
+
"pfc_gap": {
|
| 23 |
+
"max": 3.6477591193684584,
|
| 24 |
+
"mean": 0.0990505438447482,
|
| 25 |
+
"min": -4.8420685154923095,
|
| 26 |
+
"std": 0.8586369376448698
|
| 27 |
+
},
|
| 28 |
+
"pfc_generated": {
|
| 29 |
+
"max": 4.414909221516321,
|
| 30 |
+
"mean": 0.5746501914311459,
|
| 31 |
+
"min": 0.005071079598004716,
|
| 32 |
+
"std": 0.6583526046619211
|
| 33 |
+
},
|
| 34 |
+
"pfc_reference": {
|
| 35 |
+
"max": 5.707070070036645,
|
| 36 |
+
"mean": 0.4755996475863976,
|
| 37 |
+
"min": 1.734456963144955e-05,
|
| 38 |
+
"std": 0.7533267987077545
|
| 39 |
+
},
|
| 40 |
+
"right_foot_indices": [
|
| 41 |
+
11,
|
| 42 |
+
12
|
| 43 |
+
],
|
| 44 |
+
"root_index": 0,
|
| 45 |
+
"up_axis": "z"
|
| 46 |
+
},
|
| 47 |
+
"audio_metrics": {
|
| 48 |
+
"beat_align_gap": {
|
| 49 |
+
"max": 0.5701364518161618,
|
| 50 |
+
"mean": -0.06749126869399916,
|
| 51 |
+
"min": -0.7608427161001056,
|
| 52 |
+
"std": 0.18871442298460653
|
| 53 |
+
},
|
| 54 |
+
"beat_align_generated": {
|
| 55 |
+
"max": 0.9819864896355885,
|
| 56 |
+
"mean": 0.49377759948027955,
|
| 57 |
+
"min": 0.0,
|
| 58 |
+
"std": 0.14606512824675702
|
| 59 |
+
},
|
| 60 |
+
"beat_align_reference": {
|
| 61 |
+
"max": 0.9864898672266914,
|
| 62 |
+
"mean": 0.5612688681742788,
|
| 63 |
+
"min": 0.0,
|
| 64 |
+
"std": 0.13609287544141188
|
| 65 |
+
},
|
| 66 |
+
"beat_direction": "music_to_motion",
|
| 67 |
+
"beat_threshold": 0.5,
|
| 68 |
+
"frames": 60,
|
| 69 |
+
"min_motion_beat_distance_seconds": 0.25,
|
| 70 |
+
"num_samples": 512,
|
| 71 |
+
"sigma_frames": 3.0
|
| 72 |
+
},
|
| 73 |
+
"benchmark": "artifact_audio",
|
| 74 |
+
"dataset_metrics": {
|
| 75 |
+
"aioz_gdance_original_audio_test": {
|
| 76 |
+
"aistpp_edge": {
|
| 77 |
+
"div_g_gap": 0.23862403086776185,
|
| 78 |
+
"div_g_generated": 3.7473771961157687,
|
| 79 |
+
"div_g_reference": 3.508753165248007,
|
| 80 |
+
"div_k_gap": -1.4359521145663674,
|
| 81 |
+
"div_k_generated": 9.315809033822825,
|
| 82 |
+
"div_k_reference": 10.751761148389193,
|
| 83 |
+
"fid_g": 1.4318985574865248,
|
| 84 |
+
"fid_k": 7.601990475458294,
|
| 85 |
+
"geometric_feature_dim": 412,
|
| 86 |
+
"kinetic_feature_dim": 90,
|
| 87 |
+
"num_samples": 103,
|
| 88 |
+
"pfc_gap": {
|
| 89 |
+
"max": 3.2058528208797874,
|
| 90 |
+
"mean": 0.05262750505804416,
|
| 91 |
+
"min": -1.6349628441701682,
|
| 92 |
+
"std": 0.6426666823502454
|
| 93 |
+
},
|
| 94 |
+
"pfc_generated": {
|
| 95 |
+
"max": 3.4774407560932494,
|
| 96 |
+
"mean": 0.3836619604251987,
|
| 97 |
+
"min": 0.005071079598004716,
|
| 98 |
+
"std": 0.5183719007045057
|
| 99 |
+
},
|
| 100 |
+
"pfc_reference": {
|
| 101 |
+
"max": 1.8273494537376207,
|
| 102 |
+
"mean": 0.3310344553671544,
|
| 103 |
+
"min": 0.0002850814965562205,
|
| 104 |
+
"std": 0.42521448548204027
|
| 105 |
+
}
|
| 106 |
+
},
|
| 107 |
+
"audio": {
|
| 108 |
+
"beat_align_gap": {
|
| 109 |
+
"max": 0.23840448764256933,
|
| 110 |
+
"mean": -0.033628672950055886,
|
| 111 |
+
"min": -0.32453708581516766,
|
| 112 |
+
"std": 0.13742288780504877
|
| 113 |
+
},
|
| 114 |
+
"beat_align_generated": {
|
| 115 |
+
"max": 0.6725372901371526,
|
| 116 |
+
"mean": 0.4919088098323277,
|
| 117 |
+
"min": 0.24963270016294453,
|
| 118 |
+
"std": 0.09099816858378701
|
| 119 |
+
},
|
| 120 |
+
"beat_align_reference": {
|
| 121 |
+
"max": 0.723284948504687,
|
| 122 |
+
"mean": 0.5255374827823837,
|
| 123 |
+
"min": 0.33650493752041727,
|
| 124 |
+
"std": 0.08282468265444852
|
| 125 |
+
},
|
| 126 |
+
"num_samples": 103
|
| 127 |
+
},
|
| 128 |
+
"num_samples": 103,
|
| 129 |
+
"physical": {
|
| 130 |
+
"body_jerk_mean": {
|
| 131 |
+
"max": 187.8517608642578,
|
| 132 |
+
"mean": 75.8366039979805,
|
| 133 |
+
"min": 27.45358657836914,
|
| 134 |
+
"std": 24.36331538169348
|
| 135 |
+
},
|
| 136 |
+
"contact_sliding_speed": {
|
| 137 |
+
"max": 0.8738095164299011,
|
| 138 |
+
"mean": 0.3195253503915754,
|
| 139 |
+
"min": 0.043544918298721313,
|
| 140 |
+
"std": 0.16198516039471123
|
| 141 |
+
},
|
| 142 |
+
"foot_ground_error": {
|
| 143 |
+
"max": 0.08582238852977753,
|
| 144 |
+
"mean": 0.04983966779054224,
|
| 145 |
+
"min": 0.007044824305921793,
|
| 146 |
+
"std": 0.017409112859072134
|
| 147 |
+
},
|
| 148 |
+
"num_samples": 103
|
| 149 |
+
}
|
| 150 |
+
},
|
| 151 |
+
"aistpp_original_audio_test": {
|
| 152 |
+
"aistpp_edge": {
|
| 153 |
+
"div_g_gap": 0.3749239056934046,
|
| 154 |
+
"div_g_generated": 3.929406542818343,
|
| 155 |
+
"div_g_reference": 3.5544826371249383,
|
| 156 |
+
"div_k_gap": -12.850860561893567,
|
| 157 |
+
"div_k_generated": 12.909821306156998,
|
| 158 |
+
"div_k_reference": 25.760681868050565,
|
| 159 |
+
"fid_g": 2.601341460428607,
|
| 160 |
+
"fid_k": 251.18791168722862,
|
| 161 |
+
"geometric_feature_dim": 412,
|
| 162 |
+
"kinetic_feature_dim": 90,
|
| 163 |
+
"num_samples": 103,
|
| 164 |
+
"pfc_gap": {
|
| 165 |
+
"max": 2.477920920125234,
|
| 166 |
+
"mean": -0.24784057237073714,
|
| 167 |
+
"min": -4.8420685154923095,
|
| 168 |
+
"std": 1.3450989582965898
|
| 169 |
+
},
|
| 170 |
+
"pfc_generated": {
|
| 171 |
+
"max": 3.0173064117838173,
|
| 172 |
+
"mean": 0.8859167056460855,
|
| 173 |
+
"min": 0.03606813460559294,
|
| 174 |
+
"std": 0.6825868113283983
|
| 175 |
+
},
|
| 176 |
+
"pfc_reference": {
|
| 177 |
+
"max": 5.707070070036645,
|
| 178 |
+
"mean": 1.1337572780168226,
|
| 179 |
+
"min": 7.906044911863323e-05,
|
| 180 |
+
"std": 1.2009778419653923
|
| 181 |
+
}
|
| 182 |
+
},
|
| 183 |
+
"audio": {
|
| 184 |
+
"beat_align_gap": {
|
| 185 |
+
"max": 0.5701364518161618,
|
| 186 |
+
"mean": -0.0801368416977987,
|
| 187 |
+
"min": -0.7608427161001056,
|
| 188 |
+
"std": 0.28185526909472197
|
| 189 |
+
},
|
| 190 |
+
"beat_align_generated": {
|
| 191 |
+
"max": 0.9819864896355885,
|
| 192 |
+
"mean": 0.5061291397536287,
|
| 193 |
+
"min": 0.0,
|
| 194 |
+
"std": 0.2064175413162069
|
| 195 |
+
},
|
| 196 |
+
"beat_align_reference": {
|
| 197 |
+
"max": 0.9864898672266914,
|
| 198 |
+
"mean": 0.5862659814514274,
|
| 199 |
+
"min": 0.0,
|
| 200 |
+
"std": 0.19459278456334647
|
| 201 |
+
},
|
| 202 |
+
"num_samples": 103
|
| 203 |
+
},
|
| 204 |
+
"num_samples": 103,
|
| 205 |
+
"physical": {
|
| 206 |
+
"body_jerk_mean": {
|
| 207 |
+
"max": 278.0105285644531,
|
| 208 |
+
"mean": 135.99853585993202,
|
| 209 |
+
"min": 32.689208984375,
|
| 210 |
+
"std": 46.387850301317464
|
| 211 |
+
},
|
| 212 |
+
"contact_sliding_speed": {
|
| 213 |
+
"max": 1.3322360515594482,
|
| 214 |
+
"mean": 0.5159466229017499,
|
| 215 |
+
"min": 0.0,
|
| 216 |
+
"std": 0.24116125404927893
|
| 217 |
+
},
|
| 218 |
+
"foot_ground_error": {
|
| 219 |
+
"max": 0.18515852093696594,
|
| 220 |
+
"mean": 0.030137372799941057,
|
| 221 |
+
"min": 0.004513147287070751,
|
| 222 |
+
"std": 0.028280388956618472
|
| 223 |
+
},
|
| 224 |
+
"num_samples": 103
|
| 225 |
+
}
|
| 226 |
+
},
|
| 227 |
+
"compas3d_original_audio_test": {
|
| 228 |
+
"aistpp_edge": {
|
| 229 |
+
"div_g_gap": 0.17526199534596376,
|
| 230 |
+
"div_g_generated": 4.703408486137112,
|
| 231 |
+
"div_g_reference": 4.528146490791149,
|
| 232 |
+
"div_k_gap": -3.1528614038078047,
|
| 233 |
+
"div_k_generated": 8.431761165269869,
|
| 234 |
+
"div_k_reference": 11.584622569077673,
|
| 235 |
+
"fid_g": 2.805483949908389,
|
| 236 |
+
"fid_k": 67.40567433330911,
|
| 237 |
+
"geometric_feature_dim": 412,
|
| 238 |
+
"kinetic_feature_dim": 90,
|
| 239 |
+
"num_samples": 102,
|
| 240 |
+
"pfc_gap": {
|
| 241 |
+
"max": 2.932789335317619,
|
| 242 |
+
"mean": 0.15156486305032094,
|
| 243 |
+
"min": -4.195418178757522,
|
| 244 |
+
"std": 0.7163983930524794
|
| 245 |
+
},
|
| 246 |
+
"pfc_generated": {
|
| 247 |
+
"max": 3.890725147350027,
|
| 248 |
+
"mean": 0.6625580218185563,
|
| 249 |
+
"min": 0.04444435017767617,
|
| 250 |
+
"std": 0.61980077587531
|
| 251 |
+
},
|
| 252 |
+
"pfc_reference": {
|
| 253 |
+
"max": 4.452664554185064,
|
| 254 |
+
"mean": 0.5109931587682353,
|
| 255 |
+
"min": 0.030597065341000908,
|
| 256 |
+
"std": 0.5892267025941194
|
| 257 |
+
}
|
| 258 |
+
},
|
| 259 |
+
"audio": {
|
| 260 |
+
"beat_align_gap": {
|
| 261 |
+
"max": 0.20317142431931395,
|
| 262 |
+
"mean": -0.0836915032734787,
|
| 263 |
+
"min": -0.40539816524995836,
|
| 264 |
+
"std": 0.09985131455347032
|
| 265 |
+
},
|
| 266 |
+
"beat_align_generated": {
|
| 267 |
+
"max": 0.6732069018318989,
|
| 268 |
+
"mean": 0.5035843145650215,
|
| 269 |
+
"min": 0.25486845614143144,
|
| 270 |
+
"std": 0.07740773730765173
|
| 271 |
+
},
|
| 272 |
+
"beat_align_reference": {
|
| 273 |
+
"max": 0.6759029031940639,
|
| 274 |
+
"mean": 0.5872758178385001,
|
| 275 |
+
"min": 0.44007268746408307,
|
| 276 |
+
"std": 0.06101503222114459
|
| 277 |
+
},
|
| 278 |
+
"num_samples": 102
|
| 279 |
+
},
|
| 280 |
+
"num_samples": 102,
|
| 281 |
+
"physical": {
|
| 282 |
+
"body_jerk_mean": {
|
| 283 |
+
"max": 151.898193359375,
|
| 284 |
+
"mean": 93.09871916677437,
|
| 285 |
+
"min": 52.08349609375,
|
| 286 |
+
"std": 22.473365730092304
|
| 287 |
+
},
|
| 288 |
+
"contact_sliding_speed": {
|
| 289 |
+
"max": 1.037505865097046,
|
| 290 |
+
"mean": 0.49518099120434594,
|
| 291 |
+
"min": 0.1495557427406311,
|
| 292 |
+
"std": 0.18642192977168262
|
| 293 |
+
},
|
| 294 |
+
"foot_ground_error": {
|
| 295 |
+
"max": 0.05694884434342384,
|
| 296 |
+
"mean": 0.019029651656655558,
|
| 297 |
+
"min": 0.003530437359586358,
|
| 298 |
+
"std": 0.010551578513320385
|
| 299 |
+
},
|
| 300 |
+
"num_samples": 102
|
| 301 |
+
}
|
| 302 |
+
},
|
| 303 |
+
"finedance_original_audio_test": {
|
| 304 |
+
"aistpp_edge": {
|
| 305 |
+
"div_g_gap": -0.11281678280093921,
|
| 306 |
+
"div_g_generated": 4.747224781473021,
|
| 307 |
+
"div_g_reference": 4.86004156427396,
|
| 308 |
+
"div_k_gap": -5.8534723963635535,
|
| 309 |
+
"div_k_generated": 13.641405650966057,
|
| 310 |
+
"div_k_reference": 19.49487804732961,
|
| 311 |
+
"fid_g": 1.707468424788388,
|
| 312 |
+
"fid_k": 40.484632925797484,
|
| 313 |
+
"geometric_feature_dim": 412,
|
| 314 |
+
"kinetic_feature_dim": 90,
|
| 315 |
+
"num_samples": 102,
|
| 316 |
+
"pfc_gap": {
|
| 317 |
+
"max": 3.6477591193684584,
|
| 318 |
+
"mean": 0.4045132325959266,
|
| 319 |
+
"min": -2.0018882941225815,
|
| 320 |
+
"std": 0.7903098367331967
|
| 321 |
+
},
|
| 322 |
+
"pfc_generated": {
|
| 323 |
+
"max": 4.414909221516321,
|
| 324 |
+
"mean": 0.7208673930521659,
|
| 325 |
+
"min": 0.01685753904491638,
|
| 326 |
+
"std": 0.8110987203020705
|
| 327 |
+
},
|
| 328 |
+
"pfc_reference": {
|
| 329 |
+
"max": 3.122786190063731,
|
| 330 |
+
"mean": 0.3163541604562393,
|
| 331 |
+
"min": 0.0007167477980582575,
|
| 332 |
+
"std": 0.4615228956341427
|
| 333 |
+
}
|
| 334 |
+
},
|
| 335 |
+
"audio": {
|
| 336 |
+
"beat_align_gap": {
|
| 337 |
+
"max": 0.5120083942239315,
|
| 338 |
+
"mean": -0.0469387823901341,
|
| 339 |
+
"min": -0.5655078705940795,
|
| 340 |
+
"std": 0.2377869692757275
|
| 341 |
+
},
|
| 342 |
+
"beat_align_generated": {
|
| 343 |
+
"max": 0.8733484359117868,
|
| 344 |
+
"mean": 0.48259422689922377,
|
| 345 |
+
"min": 0.08184839848172423,
|
| 346 |
+
"std": 0.19826141154513752
|
| 347 |
+
},
|
| 348 |
+
"beat_align_reference": {
|
| 349 |
+
"max": 0.9155656239411911,
|
| 350 |
+
"mean": 0.529533009289358,
|
| 351 |
+
"min": 0.16565130039393888,
|
| 352 |
+
"std": 0.18269913727648307
|
| 353 |
+
},
|
| 354 |
+
"num_samples": 102
|
| 355 |
+
},
|
| 356 |
+
"num_samples": 102,
|
| 357 |
+
"physical": {
|
| 358 |
+
"body_jerk_mean": {
|
| 359 |
+
"max": 208.11827087402344,
|
| 360 |
+
"mean": 101.01181910084743,
|
| 361 |
+
"min": 15.315655708312988,
|
| 362 |
+
"std": 42.45778704969256
|
| 363 |
+
},
|
| 364 |
+
"contact_sliding_speed": {
|
| 365 |
+
"max": 1.3698070049285889,
|
| 366 |
+
"mean": 0.4525775205636141,
|
| 367 |
+
"min": 0.05821777507662773,
|
| 368 |
+
"std": 0.23398819767804652
|
| 369 |
+
},
|
| 370 |
+
"foot_ground_error": {
|
| 371 |
+
"max": 0.14215940237045288,
|
| 372 |
+
"mean": 0.03325672116011893,
|
| 373 |
+
"min": 0.0028612713795155287,
|
| 374 |
+
"std": 0.029781991560662673
|
| 375 |
+
},
|
| 376 |
+
"num_samples": 102
|
| 377 |
+
}
|
| 378 |
+
},
|
| 379 |
+
"opendance_original_audio_test": {
|
| 380 |
+
"aistpp_edge": {
|
| 381 |
+
"div_g_gap": -0.03468968675710382,
|
| 382 |
+
"div_g_generated": 4.486163913297346,
|
| 383 |
+
"div_g_reference": 4.5208536000544495,
|
| 384 |
+
"div_k_gap": -2.374843698173482,
|
| 385 |
+
"div_k_generated": 8.840375829535647,
|
| 386 |
+
"div_k_reference": 11.215219527709129,
|
| 387 |
+
"fid_g": 1.590908401644036,
|
| 388 |
+
"fid_k": 18.76958645090764,
|
| 389 |
+
"geometric_feature_dim": 412,
|
| 390 |
+
"kinetic_feature_dim": 90,
|
| 391 |
+
"num_samples": 102,
|
| 392 |
+
"pfc_gap": {
|
| 393 |
+
"max": 2.2686089657121986,
|
| 394 |
+
"mean": 0.1382437120176589,
|
| 395 |
+
"min": -0.24399738769559945,
|
| 396 |
+
"std": 0.3123090105240401
|
| 397 |
+
},
|
| 398 |
+
"pfc_generated": {
|
| 399 |
+
"max": 2.3862508697288667,
|
| 400 |
+
"mean": 0.21906767735873303,
|
| 401 |
+
"min": 0.011352475937714133,
|
| 402 |
+
"std": 0.31867181439113135
|
| 403 |
+
},
|
| 404 |
+
"pfc_reference": {
|
| 405 |
+
"max": 0.5202682826002994,
|
| 406 |
+
"mean": 0.08082396534107408,
|
| 407 |
+
"min": 1.734456963144955e-05,
|
| 408 |
+
"std": 0.09160782744166214
|
| 409 |
+
}
|
| 410 |
+
},
|
| 411 |
+
"audio": {
|
| 412 |
+
"beat_align_gap": {
|
| 413 |
+
"max": 0.21636652206933799,
|
| 414 |
+
"mean": -0.09326855318539262,
|
| 415 |
+
"min": -0.4055321230244736,
|
| 416 |
+
"std": 0.10144398731338053
|
| 417 |
+
},
|
| 418 |
+
"beat_align_generated": {
|
| 419 |
+
"max": 0.691326904650172,
|
| 420 |
+
"mean": 0.4845687342862411,
|
| 421 |
+
"min": 0.0,
|
| 422 |
+
"std": 0.09971205068090859
|
| 423 |
+
},
|
| 424 |
+
"beat_align_reference": {
|
| 425 |
+
"max": 0.7346093465204763,
|
| 426 |
+
"mean": 0.5778372874716339,
|
| 427 |
+
"min": 0.0,
|
| 428 |
+
"std": 0.08261153009190644
|
| 429 |
+
},
|
| 430 |
+
"num_samples": 102
|
| 431 |
+
},
|
| 432 |
+
"num_samples": 102,
|
| 433 |
+
"physical": {
|
| 434 |
+
"body_jerk_mean": {
|
| 435 |
+
"max": 168.04806518554688,
|
| 436 |
+
"mean": 72.5217565835691,
|
| 437 |
+
"min": 18.360395431518555,
|
| 438 |
+
"std": 24.450591057024837
|
| 439 |
+
},
|
| 440 |
+
"contact_sliding_speed": {
|
| 441 |
+
"max": 0.7384104132652283,
|
| 442 |
+
"mean": 0.24727879366015687,
|
| 443 |
+
"min": 0.0463300384581089,
|
| 444 |
+
"std": 0.12928315861184075
|
| 445 |
+
},
|
| 446 |
+
"foot_ground_error": {
|
| 447 |
+
"max": 0.09883774071931839,
|
| 448 |
+
"mean": 0.07268030654785096,
|
| 449 |
+
"min": 0.019137809053063393,
|
| 450 |
+
"std": 0.013635613677582127
|
| 451 |
+
},
|
| 452 |
+
"num_samples": 102
|
| 453 |
+
}
|
| 454 |
+
}
|
| 455 |
+
},
|
| 456 |
+
"exp": "g1_transformer_guided_300m",
|
| 457 |
+
"generated_qpos": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/generated_qpos.npz",
|
| 458 |
+
"label": "OmniMotion 300M mixed-modal s055000",
|
| 459 |
+
"num_frames": 60,
|
| 460 |
+
"num_samples": 512,
|
| 461 |
+
"physical_metrics": {
|
| 462 |
+
"body_jerk_mean": {
|
| 463 |
+
"max": 278.0105285644531,
|
| 464 |
+
"mean": 95.73342476598918,
|
| 465 |
+
"min": 15.315655708312988,
|
| 466 |
+
"std": 40.634993922345
|
| 467 |
+
},
|
| 468 |
+
"contact_height_threshold": 0.12,
|
| 469 |
+
"contact_penetration_tolerance": 0.02,
|
| 470 |
+
"contact_sliding_speed": {
|
| 471 |
+
"max": 1.3698070049285889,
|
| 472 |
+
"mean": 0.40614730156812584,
|
| 473 |
+
"min": 0.0,
|
| 474 |
+
"std": 0.2216091804538725
|
| 475 |
+
},
|
| 476 |
+
"foot_ground_error": {
|
| 477 |
+
"max": 0.18515852093696594,
|
| 478 |
+
"mean": 0.040984836867210106,
|
| 479 |
+
"min": 0.0028612713795155287,
|
| 480 |
+
"std": 0.02837769267389404
|
| 481 |
+
},
|
| 482 |
+
"frames": 60,
|
| 483 |
+
"num_samples": 512
|
| 484 |
+
},
|
| 485 |
+
"reference_qpos": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/reference_qpos.npz",
|
| 486 |
+
"sample_metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/sample_metrics.jsonl",
|
| 487 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/audio_test_512.jsonl",
|
| 488 |
+
"split": "test",
|
| 489 |
+
"tracker_executed": {
|
| 490 |
+
"action_clip": 10.0,
|
| 491 |
+
"artifact_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_executed_qpos.npz",
|
| 492 |
+
"benchmark_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/benchmark_tracker.json",
|
| 493 |
+
"control_substeps": 10,
|
| 494 |
+
"enabled": true,
|
| 495 |
+
"holomotion_onnx": "${OMG_MODELS_ROOT}/HoloMotion/HoloMotion_v1.2/holomotion_v1.2_motion_tracking_model/exported/model.onnx",
|
| 496 |
+
"metrics": {
|
| 497 |
+
"e_acc": {
|
| 498 |
+
"max": 5.082680843770504,
|
| 499 |
+
"mean": 1.897563667569102,
|
| 500 |
+
"min": 0.35615143133327365,
|
| 501 |
+
"std": 0.8002793205380141
|
| 502 |
+
},
|
| 503 |
+
"e_vel": {
|
| 504 |
+
"max": 17.04840175807476,
|
| 505 |
+
"mean": 6.103366651132092,
|
| 506 |
+
"min": 1.4394153840839863,
|
| 507 |
+
"std": 2.3867622499218415
|
| 508 |
+
},
|
| 509 |
+
"enabled": true,
|
| 510 |
+
"g_mpjpe": {
|
| 511 |
+
"max": 341.753214597702,
|
| 512 |
+
"mean": 95.1012089572032,
|
| 513 |
+
"min": 28.030861169099808,
|
| 514 |
+
"std": 39.98059037197871
|
| 515 |
+
},
|
| 516 |
+
"mpjpe": {
|
| 517 |
+
"max": 144.38872039318085,
|
| 518 |
+
"mean": 41.06933371076593,
|
| 519 |
+
"min": 18.016833811998367,
|
| 520 |
+
"std": 15.862212133962863
|
| 521 |
+
},
|
| 522 |
+
"num_samples": 512,
|
| 523 |
+
"root_index": 0,
|
| 524 |
+
"target_fps": 50.0
|
| 525 |
+
},
|
| 526 |
+
"metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_metrics.json",
|
| 527 |
+
"num_generated_samples": 512,
|
| 528 |
+
"num_samples": 512,
|
| 529 |
+
"providers": "TensorrtExecutionProvider,CUDAExecutionProvider",
|
| 530 |
+
"robot_xml": null,
|
| 531 |
+
"root_index": 0,
|
| 532 |
+
"sample_metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_sample_metrics.jsonl",
|
| 533 |
+
"selected_indices": [
|
| 534 |
+
0,
|
| 535 |
+
1,
|
| 536 |
+
2,
|
| 537 |
+
3,
|
| 538 |
+
4,
|
| 539 |
+
5,
|
| 540 |
+
6,
|
| 541 |
+
7,
|
| 542 |
+
8,
|
| 543 |
+
9,
|
| 544 |
+
10,
|
| 545 |
+
11,
|
| 546 |
+
12,
|
| 547 |
+
13,
|
| 548 |
+
14,
|
| 549 |
+
15,
|
| 550 |
+
16,
|
| 551 |
+
17,
|
| 552 |
+
18,
|
| 553 |
+
19,
|
| 554 |
+
20,
|
| 555 |
+
21,
|
| 556 |
+
22,
|
| 557 |
+
23,
|
| 558 |
+
24,
|
| 559 |
+
25,
|
| 560 |
+
26,
|
| 561 |
+
27,
|
| 562 |
+
28,
|
| 563 |
+
29,
|
| 564 |
+
30,
|
| 565 |
+
31,
|
| 566 |
+
32,
|
| 567 |
+
33,
|
| 568 |
+
34,
|
| 569 |
+
35,
|
| 570 |
+
36,
|
| 571 |
+
37,
|
| 572 |
+
38,
|
| 573 |
+
39,
|
| 574 |
+
40,
|
| 575 |
+
41,
|
| 576 |
+
42,
|
| 577 |
+
43,
|
| 578 |
+
44,
|
| 579 |
+
45,
|
| 580 |
+
46,
|
| 581 |
+
47,
|
| 582 |
+
48,
|
| 583 |
+
49,
|
| 584 |
+
50,
|
| 585 |
+
51,
|
| 586 |
+
52,
|
| 587 |
+
53,
|
| 588 |
+
54,
|
| 589 |
+
55,
|
| 590 |
+
56,
|
| 591 |
+
57,
|
| 592 |
+
58,
|
| 593 |
+
59,
|
| 594 |
+
60,
|
| 595 |
+
61,
|
| 596 |
+
62,
|
| 597 |
+
63,
|
| 598 |
+
64,
|
| 599 |
+
65,
|
| 600 |
+
66,
|
| 601 |
+
67,
|
| 602 |
+
68,
|
| 603 |
+
69,
|
| 604 |
+
70,
|
| 605 |
+
71,
|
| 606 |
+
72,
|
| 607 |
+
73,
|
| 608 |
+
74,
|
| 609 |
+
75,
|
| 610 |
+
76,
|
| 611 |
+
77,
|
| 612 |
+
78,
|
| 613 |
+
79,
|
| 614 |
+
80,
|
| 615 |
+
81,
|
| 616 |
+
82,
|
| 617 |
+
83,
|
| 618 |
+
84,
|
| 619 |
+
85,
|
| 620 |
+
86,
|
| 621 |
+
87,
|
| 622 |
+
88,
|
| 623 |
+
89,
|
| 624 |
+
90,
|
| 625 |
+
91,
|
| 626 |
+
92,
|
| 627 |
+
93,
|
| 628 |
+
94,
|
| 629 |
+
95,
|
| 630 |
+
96,
|
| 631 |
+
97,
|
| 632 |
+
98,
|
| 633 |
+
99,
|
| 634 |
+
100,
|
| 635 |
+
101,
|
| 636 |
+
102,
|
| 637 |
+
103,
|
| 638 |
+
104,
|
| 639 |
+
105,
|
| 640 |
+
106,
|
| 641 |
+
107,
|
| 642 |
+
108,
|
| 643 |
+
109,
|
| 644 |
+
110,
|
| 645 |
+
111,
|
| 646 |
+
112,
|
| 647 |
+
113,
|
| 648 |
+
114,
|
| 649 |
+
115,
|
| 650 |
+
116,
|
| 651 |
+
117,
|
| 652 |
+
118,
|
| 653 |
+
119,
|
| 654 |
+
120,
|
| 655 |
+
121,
|
| 656 |
+
122,
|
| 657 |
+
123,
|
| 658 |
+
124,
|
| 659 |
+
125,
|
| 660 |
+
126,
|
| 661 |
+
127,
|
| 662 |
+
128,
|
| 663 |
+
129,
|
| 664 |
+
130,
|
| 665 |
+
131,
|
| 666 |
+
132,
|
| 667 |
+
133,
|
| 668 |
+
134,
|
| 669 |
+
135,
|
| 670 |
+
136,
|
| 671 |
+
137,
|
| 672 |
+
138,
|
| 673 |
+
139,
|
| 674 |
+
140,
|
| 675 |
+
141,
|
| 676 |
+
142,
|
| 677 |
+
143,
|
| 678 |
+
144,
|
| 679 |
+
145,
|
| 680 |
+
146,
|
| 681 |
+
147,
|
| 682 |
+
148,
|
| 683 |
+
149,
|
| 684 |
+
150,
|
| 685 |
+
151,
|
| 686 |
+
152,
|
| 687 |
+
153,
|
| 688 |
+
154,
|
| 689 |
+
155,
|
| 690 |
+
156,
|
| 691 |
+
157,
|
| 692 |
+
158,
|
| 693 |
+
159,
|
| 694 |
+
160,
|
| 695 |
+
161,
|
| 696 |
+
162,
|
| 697 |
+
163,
|
| 698 |
+
164,
|
| 699 |
+
165,
|
| 700 |
+
166,
|
| 701 |
+
167,
|
| 702 |
+
168,
|
| 703 |
+
169,
|
| 704 |
+
170,
|
| 705 |
+
171,
|
| 706 |
+
172,
|
| 707 |
+
173,
|
| 708 |
+
174,
|
| 709 |
+
175,
|
| 710 |
+
176,
|
| 711 |
+
177,
|
| 712 |
+
178,
|
| 713 |
+
179,
|
| 714 |
+
180,
|
| 715 |
+
181,
|
| 716 |
+
182,
|
| 717 |
+
183,
|
| 718 |
+
184,
|
| 719 |
+
185,
|
| 720 |
+
186,
|
| 721 |
+
187,
|
| 722 |
+
188,
|
| 723 |
+
189,
|
| 724 |
+
190,
|
| 725 |
+
191,
|
| 726 |
+
192,
|
| 727 |
+
193,
|
| 728 |
+
194,
|
| 729 |
+
195,
|
| 730 |
+
196,
|
| 731 |
+
197,
|
| 732 |
+
198,
|
| 733 |
+
199,
|
| 734 |
+
200,
|
| 735 |
+
201,
|
| 736 |
+
202,
|
| 737 |
+
203,
|
| 738 |
+
204,
|
| 739 |
+
205,
|
| 740 |
+
206,
|
| 741 |
+
207,
|
| 742 |
+
208,
|
| 743 |
+
209,
|
| 744 |
+
210,
|
| 745 |
+
211,
|
| 746 |
+
212,
|
| 747 |
+
213,
|
| 748 |
+
214,
|
| 749 |
+
215,
|
| 750 |
+
216,
|
| 751 |
+
217,
|
| 752 |
+
218,
|
| 753 |
+
219,
|
| 754 |
+
220,
|
| 755 |
+
221,
|
| 756 |
+
222,
|
| 757 |
+
223,
|
| 758 |
+
224,
|
| 759 |
+
225,
|
| 760 |
+
226,
|
| 761 |
+
227,
|
| 762 |
+
228,
|
| 763 |
+
229,
|
| 764 |
+
230,
|
| 765 |
+
231,
|
| 766 |
+
232,
|
| 767 |
+
233,
|
| 768 |
+
234,
|
| 769 |
+
235,
|
| 770 |
+
236,
|
| 771 |
+
237,
|
| 772 |
+
238,
|
| 773 |
+
239,
|
| 774 |
+
240,
|
| 775 |
+
241,
|
| 776 |
+
242,
|
| 777 |
+
243,
|
| 778 |
+
244,
|
| 779 |
+
245,
|
| 780 |
+
246,
|
| 781 |
+
247,
|
| 782 |
+
248,
|
| 783 |
+
249,
|
| 784 |
+
250,
|
| 785 |
+
251,
|
| 786 |
+
252,
|
| 787 |
+
253,
|
| 788 |
+
254,
|
| 789 |
+
255,
|
| 790 |
+
256,
|
| 791 |
+
257,
|
| 792 |
+
258,
|
| 793 |
+
259,
|
| 794 |
+
260,
|
| 795 |
+
261,
|
| 796 |
+
262,
|
| 797 |
+
263,
|
| 798 |
+
264,
|
| 799 |
+
265,
|
| 800 |
+
266,
|
| 801 |
+
267,
|
| 802 |
+
268,
|
| 803 |
+
269,
|
| 804 |
+
270,
|
| 805 |
+
271,
|
| 806 |
+
272,
|
| 807 |
+
273,
|
| 808 |
+
274,
|
| 809 |
+
275,
|
| 810 |
+
276,
|
| 811 |
+
277,
|
| 812 |
+
278,
|
| 813 |
+
279,
|
| 814 |
+
280,
|
| 815 |
+
281,
|
| 816 |
+
282,
|
| 817 |
+
283,
|
| 818 |
+
284,
|
| 819 |
+
285,
|
| 820 |
+
286,
|
| 821 |
+
287,
|
| 822 |
+
288,
|
| 823 |
+
289,
|
| 824 |
+
290,
|
| 825 |
+
291,
|
| 826 |
+
292,
|
| 827 |
+
293,
|
| 828 |
+
294,
|
| 829 |
+
295,
|
| 830 |
+
296,
|
| 831 |
+
297,
|
| 832 |
+
298,
|
| 833 |
+
299,
|
| 834 |
+
300,
|
| 835 |
+
301,
|
| 836 |
+
302,
|
| 837 |
+
303,
|
| 838 |
+
304,
|
| 839 |
+
305,
|
| 840 |
+
306,
|
| 841 |
+
307,
|
| 842 |
+
308,
|
| 843 |
+
309,
|
| 844 |
+
310,
|
| 845 |
+
311,
|
| 846 |
+
312,
|
| 847 |
+
313,
|
| 848 |
+
314,
|
| 849 |
+
315,
|
| 850 |
+
316,
|
| 851 |
+
317,
|
| 852 |
+
318,
|
| 853 |
+
319,
|
| 854 |
+
320,
|
| 855 |
+
321,
|
| 856 |
+
322,
|
| 857 |
+
323,
|
| 858 |
+
324,
|
| 859 |
+
325,
|
| 860 |
+
326,
|
| 861 |
+
327,
|
| 862 |
+
328,
|
| 863 |
+
329,
|
| 864 |
+
330,
|
| 865 |
+
331,
|
| 866 |
+
332,
|
| 867 |
+
333,
|
| 868 |
+
334,
|
| 869 |
+
335,
|
| 870 |
+
336,
|
| 871 |
+
337,
|
| 872 |
+
338,
|
| 873 |
+
339,
|
| 874 |
+
340,
|
| 875 |
+
341,
|
| 876 |
+
342,
|
| 877 |
+
343,
|
| 878 |
+
344,
|
| 879 |
+
345,
|
| 880 |
+
346,
|
| 881 |
+
347,
|
| 882 |
+
348,
|
| 883 |
+
349,
|
| 884 |
+
350,
|
| 885 |
+
351,
|
| 886 |
+
352,
|
| 887 |
+
353,
|
| 888 |
+
354,
|
| 889 |
+
355,
|
| 890 |
+
356,
|
| 891 |
+
357,
|
| 892 |
+
358,
|
| 893 |
+
359,
|
| 894 |
+
360,
|
| 895 |
+
361,
|
| 896 |
+
362,
|
| 897 |
+
363,
|
| 898 |
+
364,
|
| 899 |
+
365,
|
| 900 |
+
366,
|
| 901 |
+
367,
|
| 902 |
+
368,
|
| 903 |
+
369,
|
| 904 |
+
370,
|
| 905 |
+
371,
|
| 906 |
+
372,
|
| 907 |
+
373,
|
| 908 |
+
374,
|
| 909 |
+
375,
|
| 910 |
+
376,
|
| 911 |
+
377,
|
| 912 |
+
378,
|
| 913 |
+
379,
|
| 914 |
+
380,
|
| 915 |
+
381,
|
| 916 |
+
382,
|
| 917 |
+
383,
|
| 918 |
+
384,
|
| 919 |
+
385,
|
| 920 |
+
386,
|
| 921 |
+
387,
|
| 922 |
+
388,
|
| 923 |
+
389,
|
| 924 |
+
390,
|
| 925 |
+
391,
|
| 926 |
+
392,
|
| 927 |
+
393,
|
| 928 |
+
394,
|
| 929 |
+
395,
|
| 930 |
+
396,
|
| 931 |
+
397,
|
| 932 |
+
398,
|
| 933 |
+
399,
|
| 934 |
+
400,
|
| 935 |
+
401,
|
| 936 |
+
402,
|
| 937 |
+
403,
|
| 938 |
+
404,
|
| 939 |
+
405,
|
| 940 |
+
406,
|
| 941 |
+
407,
|
| 942 |
+
408,
|
| 943 |
+
409,
|
| 944 |
+
410,
|
| 945 |
+
411,
|
| 946 |
+
412,
|
| 947 |
+
413,
|
| 948 |
+
414,
|
| 949 |
+
415,
|
| 950 |
+
416,
|
| 951 |
+
417,
|
| 952 |
+
418,
|
| 953 |
+
419,
|
| 954 |
+
420,
|
| 955 |
+
421,
|
| 956 |
+
422,
|
| 957 |
+
423,
|
| 958 |
+
424,
|
| 959 |
+
425,
|
| 960 |
+
426,
|
| 961 |
+
427,
|
| 962 |
+
428,
|
| 963 |
+
429,
|
| 964 |
+
430,
|
| 965 |
+
431,
|
| 966 |
+
432,
|
| 967 |
+
433,
|
| 968 |
+
434,
|
| 969 |
+
435,
|
| 970 |
+
436,
|
| 971 |
+
437,
|
| 972 |
+
438,
|
| 973 |
+
439,
|
| 974 |
+
440,
|
| 975 |
+
441,
|
| 976 |
+
442,
|
| 977 |
+
443,
|
| 978 |
+
444,
|
| 979 |
+
445,
|
| 980 |
+
446,
|
| 981 |
+
447,
|
| 982 |
+
448,
|
| 983 |
+
449,
|
| 984 |
+
450,
|
| 985 |
+
451,
|
| 986 |
+
452,
|
| 987 |
+
453,
|
| 988 |
+
454,
|
| 989 |
+
455,
|
| 990 |
+
456,
|
| 991 |
+
457,
|
| 992 |
+
458,
|
| 993 |
+
459,
|
| 994 |
+
460,
|
| 995 |
+
461,
|
| 996 |
+
462,
|
| 997 |
+
463,
|
| 998 |
+
464,
|
| 999 |
+
465,
|
| 1000 |
+
466,
|
| 1001 |
+
467,
|
| 1002 |
+
468,
|
| 1003 |
+
469,
|
| 1004 |
+
470,
|
| 1005 |
+
471,
|
| 1006 |
+
472,
|
| 1007 |
+
473,
|
| 1008 |
+
474,
|
| 1009 |
+
475,
|
| 1010 |
+
476,
|
| 1011 |
+
477,
|
| 1012 |
+
478,
|
| 1013 |
+
479,
|
| 1014 |
+
480,
|
| 1015 |
+
481,
|
| 1016 |
+
482,
|
| 1017 |
+
483,
|
| 1018 |
+
484,
|
| 1019 |
+
485,
|
| 1020 |
+
486,
|
| 1021 |
+
487,
|
| 1022 |
+
488,
|
| 1023 |
+
489,
|
| 1024 |
+
490,
|
| 1025 |
+
491,
|
| 1026 |
+
492,
|
| 1027 |
+
493,
|
| 1028 |
+
494,
|
| 1029 |
+
495,
|
| 1030 |
+
496,
|
| 1031 |
+
497,
|
| 1032 |
+
498,
|
| 1033 |
+
499,
|
| 1034 |
+
500,
|
| 1035 |
+
501,
|
| 1036 |
+
502,
|
| 1037 |
+
503,
|
| 1038 |
+
504,
|
| 1039 |
+
505,
|
| 1040 |
+
506,
|
| 1041 |
+
507,
|
| 1042 |
+
508,
|
| 1043 |
+
509,
|
| 1044 |
+
510,
|
| 1045 |
+
511
|
| 1046 |
+
],
|
| 1047 |
+
"steps": null,
|
| 1048 |
+
"target_fps": 50.0,
|
| 1049 |
+
"valid_shape": [
|
| 1050 |
+
512,
|
| 1051 |
+
99
|
| 1052 |
+
]
|
| 1053 |
+
},
|
| 1054 |
+
"transition_metrics": {
|
| 1055 |
+
"chunk_length": 60,
|
| 1056 |
+
"enabled": false,
|
| 1057 |
+
"reason": "disabled"
|
| 1058 |
+
}
|
| 1059 |
+
},
|
| 1060 |
+
"label": "OmniMotion 300M mixed-modal s055000"
|
| 1061 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/physical_metrics.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"body_jerk_mean": {
|
| 3 |
+
"max": 278.0105285644531,
|
| 4 |
+
"mean": 95.73342476598918,
|
| 5 |
+
"min": 15.315655708312988,
|
| 6 |
+
"std": 40.634993922345
|
| 7 |
+
},
|
| 8 |
+
"contact_height_threshold": 0.12,
|
| 9 |
+
"contact_penetration_tolerance": 0.02,
|
| 10 |
+
"contact_sliding_speed": {
|
| 11 |
+
"max": 1.3698070049285889,
|
| 12 |
+
"mean": 0.40614730156812584,
|
| 13 |
+
"min": 0.0,
|
| 14 |
+
"std": 0.2216091804538725
|
| 15 |
+
},
|
| 16 |
+
"foot_ground_error": {
|
| 17 |
+
"max": 0.18515852093696594,
|
| 18 |
+
"mean": 0.040984836867210106,
|
| 19 |
+
"min": 0.0028612713795155287,
|
| 20 |
+
"std": 0.02837769267389404
|
| 21 |
+
},
|
| 22 |
+
"frames": 60,
|
| 23 |
+
"num_samples": 512
|
| 24 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/sample_metrics.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_001/samples.jsonl
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 946, "index": 946, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/Ix-5MsJ2HDg_03_0_1860__person_01.json", "segment_action": "a dancer performs zumba movements to Pop music", "segment_frame_end": 1860, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Zumba", "sequence_name": "Ix-5MsJ2HDg_03_0_1860__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/Ix-5MsJ2HDg_03_0_1860__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence Ix-5MsJ2HDg_03_0_1860: Zumba to Pop music", "window_end": 960, "window_start": 900}}
|
| 2 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 959, "index": 959, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/J1sYFBv3k0w_02_764_1380__person_01.json", "segment_action": "a dancer performs commercial movements to Pop music", "segment_frame_end": 616, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "J1sYFBv3k0w_02_764_1380__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/J1sYFBv3k0w_02_764_1380__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence J1sYFBv3k0w_02_764_1380: Commercial to Pop music", "window_end": 616, "window_start": 556}}
|
| 3 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 964, "index": 964, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/J6_kvaXmv7c_03_0_1410__person_01.json", "segment_action": "a dancer performs commercial movements to Pop music", "segment_frame_end": 1410, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "J6_kvaXmv7c_03_0_1410__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/J6_kvaXmv7c_03_0_1410__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence J6_kvaXmv7c_03_0_1410: Commercial to Pop music", "window_end": 735, "window_start": 675}}
|
| 4 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 979, "index": 979, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/JAtBXSA4aLg_10_0_1830__person_02.json", "segment_action": "a dancer performs aerobic movements to Folk music", "segment_frame_end": 1830, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "JAtBXSA4aLg_10_0_1830__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/JAtBXSA4aLg_10_0_1830__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence JAtBXSA4aLg_10_0_1830: Aerobic to Folk music", "window_end": 945, "window_start": 885}}
|
| 5 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1006, "index": 1006, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/JiHRnSe2ozM_03_0_855__person_00.json", "segment_action": "a dancer performs samba movements to Latin music", "segment_frame_end": 855, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Samba", "sequence_name": "JiHRnSe2ozM_03_0_855__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/JiHRnSe2ozM_03_0_855__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence JiHRnSe2ozM_03_0_855: Samba to Latin music", "window_end": 458, "window_start": 398}}
|
| 6 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1009, "index": 1009, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/JiHRnSe2ozM_03_0_855__person_01.json", "segment_action": "a dancer performs samba movements to Latin music", "segment_frame_end": 855, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Samba", "sequence_name": "JiHRnSe2ozM_03_0_855__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/JiHRnSe2ozM_03_0_855__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence JiHRnSe2ozM_03_0_855: Samba to Latin music", "window_end": 458, "window_start": 398}}
|
| 7 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1016, "index": 1016, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/JvQHcY-KuHs_02_0_564__person_00.json", "segment_action": "a dancer performs commercial movements to R&B music", "segment_frame_end": 564, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "JvQHcY-KuHs_02_0_564__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/JvQHcY-KuHs_02_0_564__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence JvQHcY-KuHs_02_0_564: Commercial to R&B music", "window_end": 564, "window_start": 504}}
|
| 8 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1040, "index": 1040, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/O-g1bsC_bZQ_05_0_960__person_01.json", "segment_action": "a dancer performs commercial movements to Electronic music", "segment_frame_end": 960, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "O-g1bsC_bZQ_05_0_960__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/O-g1bsC_bZQ_05_0_960__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence O-g1bsC_bZQ_05_0_960: Commercial to Electronic music", "window_end": 960, "window_start": 900}}
|
| 9 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1043, "index": 1043, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/PxaGCHVh4RU_03_0_780__person_00.json", "segment_action": "a dancer performs aerobic movements to R&B music", "segment_frame_end": 780, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "PxaGCHVh4RU_03_0_780__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/PxaGCHVh4RU_03_0_780__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence PxaGCHVh4RU_03_0_780: Aerobic to R&B music", "window_end": 780, "window_start": 720}}
|
| 10 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1054, "index": 1054, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/RKKaYrn8kQM_02_0_1260__person_00.json", "segment_action": "a dancer performs commercial movements to Funk music", "segment_frame_end": 1260, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "RKKaYrn8kQM_02_0_1260__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/RKKaYrn8kQM_02_0_1260__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence RKKaYrn8kQM_02_0_1260: Commercial to Funk music", "window_end": 660, "window_start": 600}}
|
| 11 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1069, "index": 1069, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/RKKaYrn8kQM_05_0_1260__person_01.json", "segment_action": "a dancer performs commercial movements to Funk music", "segment_frame_end": 1260, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "RKKaYrn8kQM_05_0_1260__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/RKKaYrn8kQM_05_0_1260__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence RKKaYrn8kQM_05_0_1260: Commercial to Funk music", "window_end": 660, "window_start": 600}}
|
| 12 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1071, "index": 1071, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/RqgXEPVvTcs_02_0_1054__person_00.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1054, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "RqgXEPVvTcs_02_0_1054__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/RqgXEPVvTcs_02_0_1054__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence RqgXEPVvTcs_02_0_1054: Bollywood to Indian music", "window_end": 60, "window_start": 0}}
|
| 13 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1078, "index": 1078, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/RqgXEPVvTcs_02_0_1054__person_02.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1054, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "RqgXEPVvTcs_02_0_1054__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/RqgXEPVvTcs_02_0_1054__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence RqgXEPVvTcs_02_0_1054: Bollywood to Indian music", "window_end": 557, "window_start": 497}}
|
| 14 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1079, "index": 1079, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/RqgXEPVvTcs_02_0_1054__person_02.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1054, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "RqgXEPVvTcs_02_0_1054__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/RqgXEPVvTcs_02_0_1054__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence RqgXEPVvTcs_02_0_1054: Bollywood to Indian music", "window_end": 1054, "window_start": 994}}
|
| 15 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1104, "index": 1104, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/SZhU9XTCUiM_14_0_1800__person_02.json", "segment_action": "a dancer performs rumba movements to Latin music", "segment_frame_end": 1800, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "SZhU9XTCUiM_14_0_1800__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/SZhU9XTCUiM_14_0_1800__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence SZhU9XTCUiM_14_0_1800: Rumba to Latin music", "window_end": 60, "window_start": 0}}
|
| 16 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1120, "index": 1120, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/SZhU9XTCUiM_27_0_1710__person_01.json", "segment_action": "a dancer performs rumba movements to Latin music", "segment_frame_end": 1710, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "SZhU9XTCUiM_27_0_1710__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/SZhU9XTCUiM_27_0_1710__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence SZhU9XTCUiM_27_0_1710: Rumba to Latin music", "window_end": 885, "window_start": 825}}
|
| 17 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1144, "index": 1144, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/SoKZDs4EHPc_03_0_990__person_00.json", "segment_action": "a dancer performs rumba movements to Folk music", "segment_frame_end": 990, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "SoKZDs4EHPc_03_0_990__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/SoKZDs4EHPc_03_0_990__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence SoKZDs4EHPc_03_0_990: Rumba to Folk music", "window_end": 525, "window_start": 465}}
|
| 18 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1158, "index": 1158, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/SoKZDs4EHPc_04_0_1830__person_02.json", "segment_action": "a dancer performs rumba movements to Folk music", "segment_frame_end": 1830, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "SoKZDs4EHPc_04_0_1830__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/SoKZDs4EHPc_04_0_1830__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence SoKZDs4EHPc_04_0_1830: Rumba to Folk music", "window_end": 60, "window_start": 0}}
|
| 19 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1172, "index": 1172, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/agELgQzrZqw_03_0_1620__person_00.json", "segment_action": "a dancer performs aerobic movements to R&B music", "segment_frame_end": 1620, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "agELgQzrZqw_03_0_1620__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/agELgQzrZqw_03_0_1620__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence agELgQzrZqw_03_0_1620: Aerobic to R&B music", "window_end": 1620, "window_start": 1560}}
|
| 20 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1183, "index": 1183, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/agELgQzrZqw_03_0_1620__person_04.json", "segment_action": "a dancer performs aerobic movements to R&B music", "segment_frame_end": 1620, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "agELgQzrZqw_03_0_1620__person_04", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/agELgQzrZqw_03_0_1620__person_04.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence agELgQzrZqw_03_0_1620: Aerobic to R&B music", "window_end": 840, "window_start": 780}}
|
| 21 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1194, "index": 1194, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/cMA195IGdQI_05_0_780__person_01.json", "segment_action": "a dancer performs aerobic movements to Rap music", "segment_frame_end": 780, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "cMA195IGdQI_05_0_780__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/cMA195IGdQI_05_0_780__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence cMA195IGdQI_05_0_780: Aerobic to Rap music", "window_end": 60, "window_start": 0}}
|
| 22 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1204, "index": 1204, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/cMA195IGdQI_05_0_780__person_04.json", "segment_action": "a dancer performs aerobic movements to Rap music", "segment_frame_end": 780, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "cMA195IGdQI_05_0_780__person_04", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/cMA195IGdQI_05_0_780__person_04.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence cMA195IGdQI_05_0_780: Aerobic to Rap music", "window_end": 420, "window_start": 360}}
|
| 23 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1224, "index": 1224, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/dLfiEZNNMp0_02_0_1260__person_01.json", "segment_action": "a dancer performs aerobic movements to Disco music", "segment_frame_end": 1260, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "dLfiEZNNMp0_02_0_1260__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/dLfiEZNNMp0_02_0_1260__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence dLfiEZNNMp0_02_0_1260: Aerobic to Disco music", "window_end": 60, "window_start": 0}}
|
| 24 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1235, "index": 1235, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/dLfiEZNNMp0_02_0_1260__person_04.json", "segment_action": "a dancer performs aerobic movements to Disco music", "segment_frame_end": 1260, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "dLfiEZNNMp0_02_0_1260__person_04", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/dLfiEZNNMp0_02_0_1260__person_04.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence dLfiEZNNMp0_02_0_1260: Aerobic to Disco music", "window_end": 1260, "window_start": 1200}}
|
| 25 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1240, "index": 1240, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/fEwjEM5czso_02_0_1230__person_01.json", "segment_action": "a dancer performs aerobic movements to Pop music", "segment_frame_end": 1230, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "fEwjEM5czso_02_0_1230__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/fEwjEM5czso_02_0_1230__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence fEwjEM5czso_02_0_1230: Aerobic to Pop music", "window_end": 645, "window_start": 585}}
|
| 26 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1250, "index": 1250, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/fkX5vQUcQbA_04_0_1230__person_00.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1230, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "fkX5vQUcQbA_04_0_1230__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/fkX5vQUcQbA_04_0_1230__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence fkX5vQUcQbA_04_0_1230: Bollywood to Indian music", "window_end": 1230, "window_start": 1170}}
|
| 27 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1254, "index": 1254, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/fkX5vQUcQbA_04_0_1230__person_02.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1230, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "fkX5vQUcQbA_04_0_1230__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/fkX5vQUcQbA_04_0_1230__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence fkX5vQUcQbA_04_0_1230: Bollywood to Indian music", "window_end": 60, "window_start": 0}}
|
| 28 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1261, "index": 1261, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/fkX5vQUcQbA_04_0_1230__person_04.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1230, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "fkX5vQUcQbA_04_0_1230__person_04", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/fkX5vQUcQbA_04_0_1230__person_04.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence fkX5vQUcQbA_04_0_1230: Bollywood to Indian music", "window_end": 645, "window_start": 585}}
|
| 29 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1295, "index": 1295, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/fkX5vQUcQbA_19_0_1590__person_02.json", "segment_action": "a dancer performs bollywood movements to Indian music", "segment_frame_end": 1590, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Bollywood", "sequence_name": "fkX5vQUcQbA_19_0_1590__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/fkX5vQUcQbA_19_0_1590__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence fkX5vQUcQbA_19_0_1590: Bollywood to Indian music", "window_end": 1590, "window_start": 1530}}
|
| 30 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1302, "index": 1302, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/gpQkBGC4Gqg_07_0_1440__person_00.json", "segment_action": "a dancer performs commercial movements to Pop music", "segment_frame_end": 1440, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Commercial", "sequence_name": "gpQkBGC4Gqg_07_0_1440__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/gpQkBGC4Gqg_07_0_1440__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence gpQkBGC4Gqg_07_0_1440: Commercial to Pop music", "window_end": 60, "window_start": 0}}
|
| 31 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1326, "index": 1326, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/hKjvymT7Fdg_02_0_1830__person_02.json", "segment_action": "a dancer performs zumba movements to Reggae music", "segment_frame_end": 1830, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Zumba", "sequence_name": "hKjvymT7Fdg_02_0_1830__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/hKjvymT7Fdg_02_0_1830__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence hKjvymT7Fdg_02_0_1830: Zumba to Reggae music", "window_end": 60, "window_start": 0}}
|
| 32 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1341, "index": 1341, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/hlaWGB4RB08_02_0_1440__person_02.json", "segment_action": "a dancer performs aerobic movements to Electronic music", "segment_frame_end": 1440, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Aerobic", "sequence_name": "hlaWGB4RB08_02_0_1440__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/hlaWGB4RB08_02_0_1440__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence hlaWGB4RB08_02_0_1440: Aerobic to Electronic music", "window_end": 60, "window_start": 0}}
|
| 33 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1388, "index": 1388, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/jvST0TukMyE_09_0_1830__person_00.json", "segment_action": "a dancer performs rumba movements to Disco music", "segment_frame_end": 1830, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "jvST0TukMyE_09_0_1830__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/jvST0TukMyE_09_0_1830__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence jvST0TukMyE_09_0_1830: Rumba to Disco music", "window_end": 1830, "window_start": 1770}}
|
| 34 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1394, "index": 1394, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/jvST0TukMyE_09_0_1830__person_02.json", "segment_action": "a dancer performs rumba movements to Disco music", "segment_frame_end": 1830, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "jvST0TukMyE_09_0_1830__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/jvST0TukMyE_09_0_1830__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence jvST0TukMyE_09_0_1830: Rumba to Disco music", "window_end": 1830, "window_start": 1770}}
|
| 35 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1400, "index": 1400, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/jvST0TukMyE_17_0_1740__person_01.json", "segment_action": "a dancer performs rumba movements to Disco music", "segment_frame_end": 1740, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "jvST0TukMyE_17_0_1740__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/jvST0TukMyE_17_0_1740__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence jvST0TukMyE_17_0_1740: Rumba to Disco music", "window_end": 1740, "window_start": 1680}}
|
| 36 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1442, "index": 1442, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/jvST0TukMyE_33_0_1320__person_01.json", "segment_action": "a dancer performs rumba movements to Disco music", "segment_frame_end": 1320, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Rumba", "sequence_name": "jvST0TukMyE_33_0_1320__person_01", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/jvST0TukMyE_33_0_1320__person_01.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence jvST0TukMyE_33_0_1320: Rumba to Disco music", "window_end": 1320, "window_start": 1260}}
|
| 37 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1464, "index": 1464, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/l5xqQDKyKCU_17_0_900__person_00.json", "segment_action": "a dancer performs zumba movements to Pop music", "segment_frame_end": 900, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Zumba", "sequence_name": "l5xqQDKyKCU_17_0_900__person_00", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/l5xqQDKyKCU_17_0_900__person_00.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence l5xqQDKyKCU_17_0_900: Zumba to Pop music", "window_end": 60, "window_start": 0}}
|
| 38 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1506, "index": 1506, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/uitvYa5NsJI_05_0_930__person_02.json", "segment_action": "a dancer performs samba movements to R&B music", "segment_frame_end": 930, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Samba", "sequence_name": "uitvYa5NsJI_05_0_930__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/uitvYa5NsJI_05_0_930__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence uitvYa5NsJI_05_0_930: Samba to R&B music", "window_end": 60, "window_start": 0}}
|
| 39 |
+
{"caption": "", "dataset": "aioz_gdance_original_audio_test", "global_index": 1507, "index": 1507, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/labels/uitvYa5NsJI_05_0_930__person_02.json", "segment_action": "a dancer performs samba movements to R&B music", "segment_frame_end": 930, "segment_frame_start": 0, "segment_index": 0, "segment_style": "Samba", "sequence_name": "uitvYa5NsJI_05_0_930__person_02", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aioz_gdance/g1/uitvYa5NsJI_05_0_930__person_02.npz", "split": "test", "text_path": null, "video_summary": "AIOZ-GDANCE sequence uitvYa5NsJI_05_0_930: Samba to R&B music", "window_end": 495, "window_start": 435}}
|
| 40 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1537, "index": 19, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sBM_cAll_d06_mBR3_ch01_retarget__seg0000_000000_000240.json", "segment_action": "a dancer swings arms, twists torso, steps side to side with bent knees and weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gBR_sBM_cAll_d06_mBR3_ch01_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sBM_cAll_d06_mBR3_ch01_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm swings, torso twists, and side steps", "window_end": 150, "window_start": 90}}
|
| 41 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1548, "index": 30, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sBM_cAll_d06_mBR3_ch05_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends torso, swings arms, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gBR_sBM_cAll_d06_mBR3_ch05_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sBM_cAll_d06_mBR3_ch05_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with torso bends, arm swings, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 42 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1568, "index": 50, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sBM_cAll_d06_mBR4_ch05_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends forward, sweeps arms, shifts weight, and steps side to side rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic coordinated", "sequence_name": "gBR_sBM_cAll_d06_mBR4_ch05_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sBM_cAll_d06_mBR4_ch05_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic coordinated dance with forward bends, arm sweeps, and rhythmic side steps", "window_end": 240, "window_start": 180}}
|
| 43 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1571, "index": 53, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sBM_cAll_d06_mBR4_ch06_retarget__seg0000_000000_000240.json", "segment_action": "A dancer twists torso, bends knees, swings arms, and does rhythmic footwork with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "energetic hip-hop", "sequence_name": "gBR_sBM_cAll_d06_mBR4_ch06_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sBM_cAll_d06_mBR4_ch06_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "A dancer performs an energetic hip-hop sequence with dynamic twists, arm swings, and rhythmic footwork", "window_end": 240, "window_start": 180}}
|
| 44 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1594, "index": 76, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0000_000000_000240.json", "segment_action": "a dancer sways hips, swings arms, bends knees, steps side to side, spins, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with hip sways, arm gestures, spins, and dynamic footwork", "window_end": 150, "window_start": 90}}
|
| 45 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1600, "index": 82, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0002_000240_000480.json", "segment_action": "a dancer twists torso, bends forward, swings arms, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic rhythmic", "sequence_name": "gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic rhythmic sequence with torso twists, forward bends, arm swings, and side steps", "window_end": 150, "window_start": 90}}
|
| 46 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1602, "index": 84, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0003_000360_000600.json", "segment_action": "a dancer sweeps arms, twists torso, bends knees, steps side to side, and does squats", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "energetic playful", "sequence_name": "gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gBR_sFM_cAll_d06_mBR2_ch16_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an energetic playful dance with arm sweeps, torso twists, knee bends, and squats", "window_end": 60, "window_start": 0}}
|
| 47 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1636, "index": 118, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sBM_cAll_d21_mHO2_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, bends knees, swings arms, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gHO_sBM_cAll_d21_mHO2_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sBM_cAll_d21_mHO2_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with side steps, arm swings, and knee bends", "window_end": 150, "window_start": 90}}
|
| 48 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1644, "index": 126, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240.json", "segment_action": "a dancer repeatedly bends knees, sways torso, extends arms, and shifts weight side to side", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with knee bends, torso sways, and arm extensions", "window_end": 60, "window_start": 0}}
|
| 49 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1645, "index": 127, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240.json", "segment_action": "a dancer repeatedly bends knees, sways torso, extends arms, and shifts weight side to side", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sBM_cAll_d21_mHO2_ch07_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with knee bends, torso sways, and arm extensions", "window_end": 150, "window_start": 90}}
|
| 50 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1674, "index": 156, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO0_ch15_retarget__seg0006_000720_000960.json", "segment_action": "a dancer bends knees, swings arms, shifts weight, spins, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "funky coordinated energetic", "sequence_name": "gHO_sFM_cAll_d21_mHO0_ch15_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO0_ch15_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a funky coordinated dance with rhythmic knee bends, arm swings, weight shifts, and spins", "window_end": 60, "window_start": 0}}
|
| 51 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1678, "index": 160, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, twists torso, swings arms, shifts weight, and spins", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with side steps, torso twists, arm swings, and a spin", "window_end": 150, "window_start": 90}}
|
| 52 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1692, "index": 174, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0005_000600_000840.json", "segment_action": "a dancer spins, moves arms fluidly, bends knees, and shifts weight with side steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0005_000600_000840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO1_ch16_retarget__seg0005_000600_000840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with spins, arm gestures, and coordinated steps", "window_end": 60, "window_start": 0}}
|
| 53 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1720, "index": 202, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO3_ch18_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends knees, steps side to side, twists torso, and moves arms with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO3_ch18_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO3_ch18_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with side steps, torso twists, and coordinated arm movements", "window_end": 150, "window_start": 90}}
|
| 54 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1735, "index": 217, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO3_ch21_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends knees, steps side to side, swings arms, twists torso, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO3_ch21_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO3_ch21_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with knee bends, side steps, arm swings, and torso twists", "window_end": 150, "window_start": 90}}
|
| 55 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1742, "index": 224, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO4_ch19_retarget__seg0000_000000_000240.json", "segment_action": "a dancer spins, sweeps arms, bends knees, shifts weight, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic playful", "sequence_name": "gHO_sFM_cAll_d21_mHO4_ch19_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO4_ch19_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic playful sequence with spins, arm sweeps, knee bends, and rhythmic steps", "window_end": 240, "window_start": 180}}
|
| 56 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1759, "index": 241, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240.json", "segment_action": "a dancer spins repeatedly, bends knees, shifts weight, and extends arms in sync with steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic rhythmic sequence with spins, knee bends, arm extensions, and weight shifts", "window_end": 150, "window_start": 90}}
|
| 57 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1760, "index": 242, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240.json", "segment_action": "a dancer spins repeatedly, bends knees, shifts weight, and extends arms in sync with steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic rhythmic", "sequence_name": "gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gHO_sFM_cAll_d21_mHO5_ch20_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic rhythmic sequence with spins, knee bends, arm extensions, and weight shifts", "window_end": 240, "window_start": 180}}
|
| 58 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1781, "index": 263, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJB_sBM_cAll_d09_mJB2_ch02_retarget__seg0000_000000_000240.json", "segment_action": "a dancer spins, extends arms, lifts legs, and shifts weight in coordinated movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful rhythmic", "sequence_name": "gJB_sBM_cAll_d09_mJB2_ch02_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJB_sBM_cAll_d09_mJB2_ch02_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful rhythmic sequence with spins, arm extensions, and leg lifts", "window_end": 240, "window_start": 180}}
|
| 59 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1789, "index": 271, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJB_sBM_cAll_d09_mJB2_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer moves arms fluidly, sways hips, bends knees, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gJB_sBM_cAll_d09_mJB2_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJB_sBM_cAll_d09_mJB2_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm movements, hip sways, and rhythmic stepping", "window_end": 150, "window_start": 90}}
|
| 60 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1799, "index": 281, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJB_sBM_cAll_d09_mJB3_ch01_retarget__seg0000_000000_000240.json", "segment_action": "a dancer spins, moves arms fluidly, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic coordinated", "sequence_name": "gJB_sBM_cAll_d09_mJB3_ch01_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJB_sBM_cAll_d09_mJB3_ch01_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic coordinated dance with spins, arm movements, and dynamic steps", "window_end": 240, "window_start": 180}}
|
| 61 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1803, "index": 285, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends arms, spins, shifts weight, and executes fluid coordinated movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with arm extensions, spins, and fluid footwork", "window_end": 60, "window_start": 0}}
|
| 62 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1804, "index": 286, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends arms, spins, shifts weight, and executes fluid coordinated movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJB_sBM_cAll_d09_mJB3_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with arm extensions, spins, and fluid footwork", "window_end": 150, "window_start": 90}}
|
| 63 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1866, "index": 348, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sBM_cAll_d03_mJS2_ch03_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends arms, bends knees, squats low, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "gJS_sBM_cAll_d03_mJS2_ch03_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sBM_cAll_d03_mJS2_ch03_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated sequence with arm extensions, knee bends, and squats", "window_end": 60, "window_start": 0}}
|
| 64 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1922, "index": 404, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sBM_cAll_d03_mJS4_ch02_retarget__seg0000_000000_000240.json", "segment_action": "a dancer takes slow side steps, shifts weight, with arms loosely at sides", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "relaxed rhythmic", "sequence_name": "gJS_sBM_cAll_d03_mJS4_ch02_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sBM_cAll_d03_mJS4_ch02_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs relaxed rhythmic side steps with minimal upper body movement", "window_end": 240, "window_start": 180}}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_002/samples.jsonl
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1940, "index": 422, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sBM_cAll_d03_mJS4_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends arms, twists torso, steps side to side with bent knees and weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic coordinated", "sequence_name": "gJS_sBM_cAll_d03_mJS4_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sBM_cAll_d03_mJS4_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic coordinated sequence with arm extensions, torso twists, and side steps", "window_end": 240, "window_start": 180}}
|
| 2 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1952, "index": 434, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0006_000720_000960.json", "segment_action": "a dancer steps side to side, sweeps arms, bends torso, raises arms, and shifts hips rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive energetic", "sequence_name": "gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive energetic sequence with arm sweeps, torso bends, and rhythmic steps", "window_end": 240, "window_start": 180}}
|
| 3 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1954, "index": 436, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0007_000840_001080.json", "segment_action": "a dancer raises arms, bends torso, spins, steps rhythmically, and shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0007_000840_001080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS0_ch01_retarget__seg0007_000840_001080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic sequence with arm gestures, body bends, spins, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 4 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1971, "index": 453, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0002_000240_000480.json", "segment_action": "a dancer steps side to side, waves arms, raises hands, and twists torso with rhythmic shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with side steps, arm movements, and torso twists", "window_end": 60, "window_start": 0}}
|
| 5 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 1984, "index": 466, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0006_000720_000960.json", "segment_action": "a dancer rises from crouch, steps side to side, swings arms, spins, and drops to crouch", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic playful", "sequence_name": "gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS2_ch03_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic playful sequence with crouching, side steps, arm swings, and a spin", "window_end": 150, "window_start": 90}}
|
| 6 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2017, "index": 499, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS4_ch12_retarget__seg0006_000720_000960.json", "segment_action": "a dancer executes a backbend, waves arms, twists torso, and steps with rhythmic shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "gJS_sFM_cAll_d03_mJS4_ch12_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS4_ch12_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic dance with backbends, arm gestures, and rhythmic footwork", "window_end": 150, "window_start": 90}}
|
| 7 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2020, "index": 502, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends, twists torso, swings arms, sways hips, and executes coordinated turns", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic dance with torso twists, arm swings, and full-body turns", "window_end": 150, "window_start": 90}}
|
| 8 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2030, "index": 512, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0003_000360_000600.json", "segment_action": "a dancer steps side to side, twists torso, bends knees, and swings arms rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with side steps, torso twists, and knee bends", "window_end": 240, "window_start": 180}}
|
| 9 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2031, "index": 513, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0004_000480_000720.json", "segment_action": "a dancer bends forward, sways hips, shifts weight, then rises with arms raised", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0004_000480_000720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS5_ch13_retarget__seg0004_000480_000720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic sequence with torso bends, hip sways, and arm lifts", "window_end": 60, "window_start": 0}}
|
| 10 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2043, "index": 525, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gJS_sFM_cAll_d03_mJS5_ch14_retarget__seg0002_000240_000480.json", "segment_action": "a dancer waves arms, raises them overhead, twists torso, steps side to side, shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "gJS_sFM_cAll_d03_mJS5_ch14_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gJS_sFM_cAll_d03_mJS5_ch14_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic dance with arm movements, torso twists, and dynamic footwork", "window_end": 60, "window_start": 0}}
|
| 11 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2076, "index": 558, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gKR_sBM_cAll_d30_mKR2_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer alternates raising arms, leans torso, bends knees, and shifts weight side to side", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic bouncy", "sequence_name": "gKR_sBM_cAll_d30_mKR2_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gKR_sBM_cAll_d30_mKR2_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic bouncy sequence with alternating arm raises, torso leans, and side steps", "window_end": 60, "window_start": 0}}
|
| 12 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2079, "index": 561, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gKR_sBM_cAll_d30_mKR2_ch09_retarget__seg0000_000000_000240.json", "segment_action": "a dancer twists torso, uses dynamic arm gestures, steps side to side with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gKR_sBM_cAll_d30_mKR2_ch09_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gKR_sBM_cAll_d30_mKR2_ch09_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with torso twists, arm gestures, and side steps", "window_end": 60, "window_start": 0}}
|
| 13 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2108, "index": 590, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gKR_sBM_cAll_d30_mKR3_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer alternates raising arms, leans torso side to side, bends knees, shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic bouncy", "sequence_name": "gKR_sBM_cAll_d30_mKR3_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gKR_sBM_cAll_d30_mKR3_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic bouncy sequence with alternating arm raises, torso leans, and knee bends", "window_end": 240, "window_start": 180}}
|
| 14 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2179, "index": 661, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gKR_sFM_cAll_d30_mKR1_ch16_retarget__seg0003_000360_000600.json", "segment_action": "a dancer waves arms, twists torso, steps side to side, spins, and shifts weight dynamically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gKR_sFM_cAll_d30_mKR1_ch16_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gKR_sFM_cAll_d30_mKR1_ch16_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with arm gestures, torso twists, spins, and dynamic footwork", "window_end": 150, "window_start": 90}}
|
| 15 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2241, "index": 723, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gKR_sFM_cAll_d30_mKR3_ch21_retarget__seg0003_000360_000600.json", "segment_action": "a dancer swings arms, twists torso, sways hips, and shifts feet with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "sharp rhythmic hip-hop", "sequence_name": "gKR_sFM_cAll_d30_mKR3_ch21_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gKR_sFM_cAll_d30_mKR3_ch21_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a sharp rhythmic hip-hop sequence with arm swings, torso twists, and dynamic footwork", "window_end": 60, "window_start": 0}}
|
| 16 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2318, "index": 800, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sBM_cAll_d18_mLH3_ch01_retarget__seg0000_000000_000240.json", "segment_action": "a dancer makes arm gestures, twists torso, steps side to side, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sBM_cAll_d18_mLH3_ch01_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sBM_cAll_d18_mLH3_ch01_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso twists, and side steps", "window_end": 240, "window_start": 180}}
|
| 17 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2323, "index": 805, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sBM_cAll_d18_mLH3_ch03_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, swings arms, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic groovy", "sequence_name": "gLH_sBM_cAll_d18_mLH3_ch03_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sBM_cAll_d18_mLH3_ch03_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic groovy sequence with side steps, arm swings, and torso twists", "window_end": 150, "window_start": 90}}
|
| 18 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2338, "index": 820, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sBM_cAll_d18_mLH3_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer sways hips, twists torso, moves arms rhythmically, and shifts weight with small steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sBM_cAll_d18_mLH3_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sBM_cAll_d18_mLH3_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with hip sways, torso twists, arm gestures, and weight shifts", "window_end": 150, "window_start": 90}}
|
| 19 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2345, "index": 827, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sBM_cAll_d18_mLH3_ch10_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, swings arms, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sBM_cAll_d18_mLH3_ch10_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sBM_cAll_d18_mLH3_ch10_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with side steps, arm swings, and body sways", "window_end": 240, "window_start": 180}}
|
| 20 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2355, "index": 837, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sBM_cAll_d18_mLH4_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, twists torso, bends knees, and uses sweeping arm movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sBM_cAll_d18_mLH4_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sBM_cAll_d18_mLH4_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with side steps, torso twists, knee bends, and sweeping arm gestures", "window_end": 60, "window_start": 0}}
|
| 21 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2399, "index": 881, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0007_000840_001080.json", "segment_action": "a dancer moves arms dynamically, twists torso, sways hips, and steps with rhythmic weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0007_000840_001080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0007_000840_001080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, hip sways, and coordinated steps", "window_end": 240, "window_start": 180}}
|
| 22 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2403, "index": 885, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0009_001080_001320.json", "segment_action": "a dancer twists torso, uses varied arm gestures, steps side to side, turns body, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic energetic", "sequence_name": "gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH0_ch15_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with torso twists, arm gestures, side steps, and body turns", "window_end": 60, "window_start": 0}}
|
| 23 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2465, "index": 947, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH2_ch17_retarget__seg0006_000720_000960.json", "segment_action": "a dancer sways hips, twists torso, steps side to side, and makes coordinated arm movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sFM_cAll_d18_mLH2_ch17_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH2_ch17_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with hip sways, torso twists, side steps, and arm gestures", "window_end": 240, "window_start": 180}}
|
| 24 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2471, "index": 953, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0000_000000_000240.json", "segment_action": "a dancer moves arms dynamically, sways torso, steps side to side, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso sways, and side steps", "window_end": 240, "window_start": 180}}
|
| 25 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2477, "index": 959, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0002_000240_000480.json", "segment_action": "a dancer steps side to side, swings arms, bends torso, shifts weight, and extends arms", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic expressive", "sequence_name": "gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH3_ch18_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic expressive sequence with coordinated steps, arm gestures, and torso bends", "window_end": 240, "window_start": 180}}
|
| 26 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2497, "index": 979, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLH_sFM_cAll_d18_mLH4_ch19_retarget__seg0002_000240_000480.json", "segment_action": "a dancer waves arms, sways hips, jumps lightly, steps side to side, twists torso", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gLH_sFM_cAll_d18_mLH4_ch19_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLH_sFM_cAll_d18_mLH4_ch19_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic sequence with arm movements, hip sways, jumps, and side steps", "window_end": 150, "window_start": 90}}
|
| 27 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2536, "index": 1018, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sBM_cAll_d15_mLO2_ch03_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends one arm, twists torso, bends knees, and rotates body repeatedly", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "sharp coordinated", "sequence_name": "gLO_sBM_cAll_d15_mLO2_ch03_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sBM_cAll_d15_mLO2_ch03_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a sharp coordinated sequence with arm extensions, torso twists, and body turns", "window_end": 150, "window_start": 90}}
|
| 28 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2594, "index": 1076, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sBM_cAll_d15_mLO4_ch02_retarget__seg0000_000000_000240.json", "segment_action": "a dancer alternates arm waves, bends torso, tilts head, and shifts weight subtly", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "gLO_sBM_cAll_d15_mLO4_ch02_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sBM_cAll_d15_mLO4_ch02_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic sequence with arm waves, torso bends, and head tilts", "window_end": 240, "window_start": 180}}
|
| 29 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2600, "index": 1082, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sBM_cAll_d15_mLO4_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends an open hand forward, bends knees, shifts hips, and repeats the motion", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gLO_sBM_cAll_d15_mLO4_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sBM_cAll_d15_mLO4_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic sequence with arm gestures, knee bends, and hip shifts", "window_end": 240, "window_start": 180}}
|
| 30 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2601, "index": 1083, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sBM_cAll_d15_mLO4_ch05_retarget__seg0000_000000_000240.json", "segment_action": "a dancer extends arms, places hand on hip, sways hips, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLO_sBM_cAll_d15_mLO4_ch05_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sBM_cAll_d15_mLO4_ch05_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, hip sways, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 31 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2608, "index": 1090, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sBM_cAll_d15_mLO4_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends knees, taps feet rhythmically, sways torso, and moves arms in sync", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "bouncy rhythmic", "sequence_name": "gLO_sBM_cAll_d15_mLO4_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sBM_cAll_d15_mLO4_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a bouncy rhythmic sequence with knee bends, foot taps, and coordinated arm movements", "window_end": 150, "window_start": 90}}
|
| 32 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2655, "index": 1137, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO1_ch16_retarget__seg0003_000360_000600.json", "segment_action": "a dancer bends knees, swings arms, twists torso, shifts weight, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO1_ch16_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO1_ch16_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with knee bends, arm gestures, torso twists, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 33 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2683, "index": 1165, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0003_000360_000600.json", "segment_action": "a dancer steps side to side, swings arms, twists torso, and bends knees rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful bouncy", "sequence_name": "gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful bouncy dance with rhythmic steps, arm swings, and torso twists", "window_end": 150, "window_start": 90}}
|
| 34 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2685, "index": 1167, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0004_000480_000720.json", "segment_action": "a dancer spins, lifts knees, waves arms, sways torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated energetic", "sequence_name": "gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0004_000480_000720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0004_000480_000720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with spins, arm waves, knee lifts, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 35 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2688, "index": 1170, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0005_000600_000840.json", "segment_action": "a dancer waves arms, twists torso, bends knees, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0005_000600_000840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO2_ch17_retarget__seg0005_000600_000840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with arm gestures, torso twists, and dynamic footwork", "window_end": 60, "window_start": 0}}
|
| 36 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2715, "index": 1197, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO3_ch18_retarget__seg0006_000720_000960.json", "segment_action": "a dancer twists torso, waves arms, steps side to side, and does sharp leg kicks with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy energetic rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO3_ch18_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO3_ch18_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an energetic groovy dance with torso twists, arm gestures, leg kicks, and rhythmic weight shifts", "window_end": 60, "window_start": 0}}
|
| 37 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2742, "index": 1224, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO4_ch21_retarget__seg0001_000120_000360.json", "segment_action": "a dancer steps side to side, sways hips, uses arm gestures, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO4_ch21_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO4_ch21_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic sequence with side steps, hip sways, arm gestures, and torso twists", "window_end": 60, "window_start": 0}}
|
| 38 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2760, "index": 1242, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0000_000000_000240.json", "segment_action": "a dancer moves arms dynamically, bends torso, steps side to side, sways hips, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso bends, side steps, and hip sways", "window_end": 60, "window_start": 0}}
|
| 39 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2764, "index": 1246, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360.json", "segment_action": "a dancer waves arms, bends knees, steps side to side, spins, and shifts weight dynamically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, knee bends, spins, and dynamic footwork", "window_end": 150, "window_start": 90}}
|
| 40 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2765, "index": 1247, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360.json", "segment_action": "a dancer waves arms, bends knees, steps side to side, spins, and shifts weight dynamically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gLO_sFM_cAll_d15_mLO5_ch20_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, knee bends, spins, and dynamic footwork", "window_end": 240, "window_start": 180}}
|
| 41 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2814, "index": 1296, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sBM_cAll_d24_mMH3_ch03_retarget__seg0000_000000_000240.json", "segment_action": "a dancer twists torso, swings arms, bends knees, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gMH_sBM_cAll_d24_mMH3_ch03_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sBM_cAll_d24_mMH3_ch03_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with torso twists, arm swings, and coordinated footwork", "window_end": 60, "window_start": 0}}
|
| 42 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2829, "index": 1311, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sBM_cAll_d24_mMH3_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, swings arms alternately, twists torso, and shifts hips with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gMH_sBM_cAll_d24_mMH3_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sBM_cAll_d24_mMH3_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with side steps, arm swings, torso twists, and hip shifts", "window_end": 60, "window_start": 0}}
|
| 43 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2887, "index": 1369, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH0_ch15_retarget__seg0006_000720_000960.json", "segment_action": "A dancer sweeps arms, twists torso, bends knees, shifts weight, and executes dynamic footwork", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "funky coordinated energetic", "sequence_name": "gMH_sFM_cAll_d24_mMH0_ch15_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH0_ch15_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "A dancer performs a funky coordinated sequence with dynamic arm gestures, torso twists, and rhythmic footwork", "window_end": 150, "window_start": 90}}
|
| 44 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2906, "index": 1388, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0002_000240_000480.json", "segment_action": "a dancer bends torso, swings arms, steps side to side, twists hips, and makes playful gestures", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with torso bends, arm swings, and hip twists", "window_end": 240, "window_start": 180}}
|
| 45 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2907, "index": 1389, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0003_000360_000600.json", "segment_action": "a dancer steps side to side, twists torso, moves arms, sways hips, and shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic groovy", "sequence_name": "gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH1_ch16_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic groovy hip-hop sequence with side steps, torso twists, and dynamic arm gestures", "window_end": 60, "window_start": 0}}
|
| 46 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2964, "index": 1446, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH2_ch21_retarget__seg0005_000600_000840.json", "segment_action": "a dancer sweeps arms, spins, lifts legs, bends torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "gMH_sFM_cAll_d24_mMH2_ch21_retarget__seg0005_000600_000840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH2_ch21_retarget__seg0005_000600_000840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic dance with arm sweeps, spins, leg lifts, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 47 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 2985, "index": 1467, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH3_ch18_retarget__seg0004_000480_000720.json", "segment_action": "a dancer steps side to side, twists torso, swings arms, and shifts weight with bent knees", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gMH_sFM_cAll_d24_mMH3_ch18_retarget__seg0004_000480_000720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH3_ch18_retarget__seg0004_000480_000720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with side steps, torso twists, and arm swings", "window_end": 60, "window_start": 0}}
|
| 48 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3002, "index": 1484, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH4_ch19_retarget__seg0002_000240_000480.json", "segment_action": "a dancer spins, swings arms, bends torso, lifts knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "energetic sharp", "sequence_name": "gMH_sFM_cAll_d24_mMH4_ch19_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH4_ch19_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an energetic sharp dance with spins, arm swings, torso bends, and dynamic footwork", "window_end": 240, "window_start": 180}}
|
| 49 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3025, "index": 1507, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0003_000360_000600.json", "segment_action": "a dancer twists torso, bends knees, moves arms, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful dynamic coordinated", "sequence_name": "gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful dynamic dance with arm movements, torso twists, knee bends, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 50 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3028, "index": 1510, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0004_000480_000720.json", "segment_action": "a dancer swings arms, twists torso, bends knees, steps side to side, and spins with coordination", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic dynamic", "sequence_name": "gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0004_000480_000720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gMH_sFM_cAll_d24_mMH5_ch20_retarget__seg0004_000480_000720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm swings, torso twists, and dynamic weight shifts", "window_end": 150, "window_start": 90}}
|
| 51 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3062, "index": 1544, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sBM_cAll_d12_mPO2_ch10_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, twists torso, moves arms in sync, and returns to stance", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "smooth coordinated", "sequence_name": "gPO_sBM_cAll_d12_mPO2_ch10_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sBM_cAll_d12_mPO2_ch10_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a smooth coordinated sequence with side steps, torso twists, and synchronized arm movements", "window_end": 240, "window_start": 180}}
|
| 52 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3073, "index": 1555, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sBM_cAll_d12_mPO3_ch04_retarget__seg0000_000000_000240.json", "segment_action": "a dancer twists torso, moves arms rhythmically, shifts weight, and turns body between front and side", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic coordinated", "sequence_name": "gPO_sBM_cAll_d12_mPO3_ch04_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sBM_cAll_d12_mPO3_ch04_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic coordinated sequence with torso twists, arm movements, and weight shifts", "window_end": 150, "window_start": 90}}
|
| 53 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3077, "index": 1559, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sBM_cAll_d12_mPO3_ch05_retarget__seg0000_000000_000240.json", "segment_action": "a dancer waves arms, twists torso, turns body, bends knees, and sways hips rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gPO_sBM_cAll_d12_mPO3_ch05_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sBM_cAll_d12_mPO3_ch05_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm waves, torso twists, knee bends, and hip sways", "window_end": 240, "window_start": 180}}
|
| 54 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3092, "index": 1574, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sBM_cAll_d12_mPO3_ch10_retarget__seg0000_000000_000240.json", "segment_action": "a dancer bends knees, steps side to side, twists torso, and moves arms in sync", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic coordinated", "sequence_name": "gPO_sBM_cAll_d12_mPO3_ch10_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sBM_cAll_d12_mPO3_ch10_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic coordinated sequence with side steps, knee bends, and torso twists", "window_end": 240, "window_start": 180}}
|
| 55 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3114, "index": 1596, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sBM_cAll_d12_mPO4_ch08_retarget__seg0000_000000_000240.json", "segment_action": "a dancer steps side to side, twists torso, bends knees, and moves arms rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic coordinated", "sequence_name": "gPO_sBM_cAll_d12_mPO4_ch08_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sBM_cAll_d12_mPO4_ch08_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic coordinated sequence with side steps, torso twists, and arm movements", "window_end": 60, "window_start": 0}}
|
| 56 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3129, "index": 1611, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480.json", "segment_action": "a dancer kicks legs, waves arms, twists torso, and shifts weight in coordinated steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with leg kicks, arm gestures, torso twists, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 57 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3130, "index": 1612, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480.json", "segment_action": "a dancer kicks legs, waves arms, twists torso, and shifts weight in coordinated steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with leg kicks, arm gestures, torso twists, and weight shifts", "window_end": 150, "window_start": 90}}
|
| 58 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3142, "index": 1624, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0006_000720_000960.json", "segment_action": "a dancer swings arms, twists torso, bends knees, steps side to side, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful groovy", "sequence_name": "gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO0_ch15_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful groovy sequence with arm swings, torso twists, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 59 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3159, "index": 1641, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO1_ch16_retarget__seg0002_000240_000480.json", "segment_action": "a dancer waves arms, twists torso, steps side to side, and shifts weight with hand gestures", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "gPO_sFM_cAll_d12_mPO1_ch16_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO1_ch16_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with arm waves, torso twists, and rhythmic stepping", "window_end": 60, "window_start": 0}}
|
| 60 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3184, "index": 1666, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0001_000120_000360.json", "segment_action": "a dancer makes arm gestures, lifts knees, steps side to side, sways torso, and shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with arm gestures, knee lifts, side steps, and body sways", "window_end": 150, "window_start": 90}}
|
| 61 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3197, "index": 1679, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0005_000600_000840.json", "segment_action": "a dancer turns, extends arms, bends torso, and steps side to side with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0005_000600_000840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO2_ch17_retarget__seg0005_000600_000840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso bends, and side steps", "window_end": 240, "window_start": 180}}
|
| 62 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3218, "index": 1700, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO3_ch18_retarget__seg0004_000480_000720.json", "segment_action": "a dancer bends knees, sweeps arms, leans torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO3_ch18_retarget__seg0004_000480_000720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO3_ch18_retarget__seg0004_000480_000720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with knee bends, arm sweeps, and torso leans", "window_end": 240, "window_start": 180}}
|
| 63 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3233, "index": 1715, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0002_000240_000480.json", "segment_action": "a dancer moves arms, sways hips, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, hip sways, knee bends, and weight shifts", "window_end": 240, "window_start": 180}}
|
| 64 |
+
{"caption": "", "dataset": "aistpp_original_audio_test", "global_index": 3235, "index": 1717, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/labels/gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0003_000360_000600.json", "segment_action": "a dancer twists torso, swings arms, steps side to side, and shifts weight with hip sways", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/aistpp/g1/gPO_sFM_cAll_d12_mPO4_ch19_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with torso twists, arm swings, and side steps", "window_end": 150, "window_start": 90}}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_004/samples.jsonl
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4258, "index": 539, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0003_00006440_00008920.json", "segment_action": "a leader performs salsa dance: Dile que no with normal closed hold, both hands", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0003_00006440_00008920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0003_00006440_00008920.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Dile que no with normal closed hold, both hands", "window_end": 68, "window_start": 8}}
|
| 2 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4275, "index": 556, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0009_00021640_00024160.json", "segment_action": "a leader performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0009_00021640_00024160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0009_00021640_00024160.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Enchufla with normal open hold", "window_end": 60, "window_start": 0}}
|
| 3 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4278, "index": 559, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0010_00024160_00026600.json", "segment_action": "a leader performs salsa dance: Dile que no with normal closed hold, both hands", "segment_frame_end": 73, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0010_00024160_00026600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0010_00024160_00026600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Dile que no with normal closed hold, both hands", "window_end": 60, "window_start": 0}}
|
| 4 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4279, "index": 560, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0010_00024160_00026600.json", "segment_action": "a leader performs salsa dance: Dile que no with normal closed hold, both hands", "segment_frame_end": 73, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0010_00024160_00026600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0010_00024160_00026600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Dile que no with normal closed hold, both hands", "window_end": 66, "window_start": 6}}
|
| 5 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4314, "index": 595, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0022_00054400_00056920.json", "segment_action": "a leader performs salsa dance: XBL with normal open hold", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0022_00054400_00056920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0022_00054400_00056920.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: XBL with normal open hold", "window_end": 60, "window_start": 0}}
|
| 6 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4322, "index": 603, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0024_00059480_00062000.json", "segment_action": "a leader performs salsa dance: XBL with normal open hold", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0024_00059480_00062000", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0024_00059480_00062000.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: XBL with normal open hold", "window_end": 76, "window_start": 16}}
|
| 7 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4333, "index": 614, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0028_00069600_00072040.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 73, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0028_00069600_00072040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0028_00069600_00072040.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Basic step, facing the line of dance", "window_end": 66, "window_start": 6}}
|
| 8 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4334, "index": 615, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0028_00069600_00072040.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 73, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0028_00069600_00072040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0028_00069600_00072040.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Basic step, facing the line of dance", "window_end": 73, "window_start": 13}}
|
| 9 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4371, "index": 652, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0041_00102280_00104830.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, normal open hold", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0041_00102280_00104830", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0041_00102280_00104830.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Right turn for the follower, normal open hold", "window_end": 60, "window_start": 0}}
|
| 10 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4379, "index": 660, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0043_00107400_00109880.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, normal open hold, ends with Left HL for the follower", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0043_00107400_00109880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0043_00107400_00109880.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Right turn for the follower, normal open hold, ends with Left HL for the follower", "window_end": 74, "window_start": 14}}
|
| 11 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4407, "index": 688, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0054_00135120_00137600.json", "segment_action": "a leader performs salsa dance: Enchufla with crossed hold to open hold while hands change", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0054_00135120_00137600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0054_00135120_00137600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Enchufla with crossed hold to open hold while hands change", "window_end": 60, "window_start": 0}}
|
| 12 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4415, "index": 696, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0056_00140120_00142600.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, right turn for the leader, normal open hold, hands change over the follower and then behind the leader", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0056_00140120_00142600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0056_00140120_00142600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Right turn for the follower, right turn for the leader, normal open hold, hands change over the follower and then behind the leader", "window_end": 74, "window_start": 14}}
|
| 13 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4418, "index": 699, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0057_00142600_00145240.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, crossed hold", "segment_frame_end": 79, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0057_00142600_00145240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0057_00142600_00145240.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Right turn for the follower, crossed hold", "window_end": 79, "window_start": 19}}
|
| 14 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4424, "index": 705, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair3_song1_take2__leader__seg0059_00147680_00150150.json", "segment_action": "a leader performs salsa dance: Copa with the follower's right (outside) turn with crossed hold", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair3_song1_take2__leader__seg0059_00147680_00150150", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair3_song1_take2__leader__seg0059_00147680_00150150.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair3/Pair3_song1_take2 leader: Copa with the follower's right (outside) turn with crossed hold", "window_end": 74, "window_start": 14}}
|
| 15 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4429, "index": 710, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0002_00003620_00006140.json", "segment_action": "a follower performs salsa dance: XBL with normal open hold", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0002_00003620_00006140", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0002_00003620_00006140.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: XBL with normal open hold", "window_end": 60, "window_start": 0}}
|
| 16 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4440, "index": 721, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0005_00011140_00013660.json", "segment_action": "a follower performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0005_00011140_00013660", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0005_00011140_00013660.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Enchufla with normal open hold", "window_end": 76, "window_start": 16}}
|
| 17 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4445, "index": 726, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0007_00016219_00018700.json", "segment_action": "a follower performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0007_00016219_00018700", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0007_00016219_00018700.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Basic step, facing the line of dance", "window_end": 67, "window_start": 7}}
|
| 18 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4457, "index": 738, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0011_00026220_00028820.json", "segment_action": "a follower performs salsa dance: Dile que no with normal open hold", "segment_frame_end": 78, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0011_00026220_00028820", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0011_00026220_00028820.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Dile que no with normal open hold", "window_end": 69, "window_start": 9}}
|
| 19 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4466, "index": 747, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0014_00033860_00036420.json", "segment_action": "a follower performs salsa dance: Copa with the follower's left (inside) turn by holding the follower's hip", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0014_00033860_00036420", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0014_00033860_00036420.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Copa with the follower's left (inside) turn by holding the follower's hip", "window_end": 68, "window_start": 8}}
|
| 20 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4470, "index": 751, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0016_00038980_00041650.json", "segment_action": "a follower performs salsa dance: Right turn for the follower, normal open hold, ends with Left HL for the follower", "segment_frame_end": 81, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0016_00038980_00041650", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0016_00038980_00041650.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Right turn for the follower, normal open hold, ends with Left HL for the follower", "window_end": 81, "window_start": 21}}
|
| 21 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4480, "index": 761, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0020_00049040_00051600.json", "segment_action": "a follower performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0020_00049040_00051600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0020_00049040_00051600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Basic step, facing the line of dance", "window_end": 60, "window_start": 0}}
|
| 22 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4501, "index": 782, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0027_00066640_00069200.json", "segment_action": "a follower performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0027_00066640_00069200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0027_00066640_00069200.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Enchufla with normal open hold", "window_end": 60, "window_start": 0}}
|
| 23 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4504, "index": 785, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0028_00069200_00071680.json", "segment_action": "a follower performs salsa dance: Dile que no with normal closed hold", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0028_00069200_00071680", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0028_00069200_00071680.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Dile que no with normal closed hold", "window_end": 60, "window_start": 0}}
|
| 24 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4507, "index": 788, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0029_00071680_00074200.json", "segment_action": "a follower performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0029_00071680_00074200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0029_00071680_00074200.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Basic step, facing the line of dance", "window_end": 60, "window_start": 0}}
|
| 25 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4545, "index": 826, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0041_00102000_00104600.json", "segment_action": "a follower performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 78, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0041_00102000_00104600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0041_00102000_00104600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Basic step, facing the line of dance", "window_end": 78, "window_start": 18}}
|
| 26 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4553, "index": 834, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0044_00119560_00122200.json", "segment_action": "a follower performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 79, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0044_00119560_00122200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0044_00119560_00122200.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Basic step, facing the line of dance", "window_end": 70, "window_start": 10}}
|
| 27 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4576, "index": 857, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0052_00139840_00142360.json", "segment_action": "a follower performs salsa dance: Enchufla from left HL for the follower with left turn for the leader", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0052_00139840_00142360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0052_00139840_00142360.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Enchufla from left HL for the follower with left turn for the leader", "window_end": 60, "window_start": 0}}
|
| 28 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4579, "index": 860, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0053_00142360_00144880.json", "segment_action": "a follower performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0053_00142360_00144880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0053_00142360_00144880.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Enchufla with normal open hold", "window_end": 60, "window_start": 0}}
|
| 29 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4580, "index": 861, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0053_00142360_00144880.json", "segment_action": "a follower performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0053_00142360_00144880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0053_00142360_00144880.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Enchufla with normal open hold", "window_end": 68, "window_start": 8}}
|
| 30 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4596, "index": 877, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0058_00033860_00036420.json", "segment_action": "a follower performs salsa dance: Lady styling", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0058_00033860_00036420", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0058_00033860_00036420.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Lady styling", "window_end": 77, "window_start": 17}}
|
| 31 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4604, "index": 885, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0062_00059080_00061600.json", "segment_action": "a follower performs salsa dance: Lady styling", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0062_00059080_00061600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0062_00059080_00061600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Lady styling", "window_end": 68, "window_start": 8}}
|
| 32 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4612, "index": 893, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0065_00107120_00109520.json", "segment_action": "a follower performs salsa dance: Steps sideways on the beat 1 and 5 on the same spot", "segment_frame_end": 72, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0065_00107120_00109520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0065_00107120_00109520.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Steps sideways on the beat 1 and 5 on the same spot", "window_end": 60, "window_start": 0}}
|
| 33 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4617, "index": 898, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0066_00109520_00112080.json", "segment_action": "a follower performs salsa dance: Steps sideways on the beat 1 and 5 on the same spot", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0066_00109520_00112080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0066_00109520_00112080.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Steps sideways on the beat 1 and 5 on the same spot", "window_end": 76, "window_start": 16}}
|
| 34 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4621, "index": 902, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__follower__seg0068_00114640_00117080.json", "segment_action": "a follower performs salsa dance: Steps sideways on the heels on the beat 5 and 7", "segment_frame_end": 73, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__follower__seg0068_00114640_00117080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__follower__seg0068_00114640_00117080.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 follower: Steps sideways on the heels on the beat 5 and 7", "window_end": 60, "window_start": 0}}
|
| 35 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4629, "index": 910, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0002_00003620_00006140.json", "segment_action": "a leader performs salsa dance: XBL with normal open hold", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0002_00003620_00006140", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0002_00003620_00006140.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: XBL with normal open hold", "window_end": 68, "window_start": 8}}
|
| 36 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4639, "index": 920, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0005_00011140_00013660.json", "segment_action": "a leader performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0005_00011140_00013660", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0005_00011140_00013660.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Enchufla with normal open hold", "window_end": 76, "window_start": 16}}
|
| 37 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4656, "index": 937, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0011_00026220_00028820.json", "segment_action": "a leader performs salsa dance: Dile que no with normal open hold", "segment_frame_end": 78, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0011_00026220_00028820", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0011_00026220_00028820.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Dile que no with normal open hold", "window_end": 69, "window_start": 9}}
|
| 38 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4662, "index": 943, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0013_00031340_00033860.json", "segment_action": "a leader performs salsa dance: Check by holding the follower's hand (open hands down) after XBL with a left (inside) turn for the follower.", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0013_00031340_00033860", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0013_00031340_00033860.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Check by holding the follower's hand (open hands down) after XBL with a left (inside) turn for the follower.", "window_end": 68, "window_start": 8}}
|
| 39 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4663, "index": 944, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0013_00031340_00033860.json", "segment_action": "a leader performs salsa dance: Check by holding the follower's hand (open hands down) after XBL with a left (inside) turn for the follower.", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0013_00031340_00033860", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0013_00031340_00033860.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Check by holding the follower's hand (open hands down) after XBL with a left (inside) turn for the follower.", "window_end": 76, "window_start": 16}}
|
| 40 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4672, "index": 953, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0016_00038980_00041650.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, normal open hold, ends with Left HL for the follower", "segment_frame_end": 81, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0016_00038980_00041650", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0016_00038980_00041650.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Right turn for the follower, normal open hold, ends with Left HL for the follower", "window_end": 81, "window_start": 21}}
|
| 41 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4678, "index": 959, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0018_00044060_00046660.json", "segment_action": "a leader performs salsa dance: Enchufla with normal open hold", "segment_frame_end": 78, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0018_00044060_00046660", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0018_00044060_00046660.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Enchufla with normal open hold", "window_end": 78, "window_start": 18}}
|
| 42 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4684, "index": 965, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0020_00049040_00051600.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0020_00049040_00051600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0020_00049040_00051600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 77, "window_start": 17}}
|
| 43 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4710, "index": 991, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0029_00071680_00074200.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0029_00071680_00074200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0029_00071680_00074200.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 68, "window_start": 8}}
|
| 44 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4720, "index": 1001, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0032_00079240_00081760.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0032_00079240_00081760", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0032_00079240_00081760.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 76, "window_start": 16}}
|
| 45 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4736, "index": 1017, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0038_00094400_00096880.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 74, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0038_00094400_00096880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0038_00094400_00096880.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 60, "window_start": 0}}
|
| 46 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4745, "index": 1026, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0041_00102000_00104600.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 78, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0041_00102000_00104600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0041_00102000_00104600.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 60, "window_start": 0}}
|
| 47 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4752, "index": 1033, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0043_00117080_00119560.json", "segment_action": "a leader performs salsa dance: XBL with normal closed hold", "segment_frame_end": 75, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0043_00117080_00119560", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0043_00117080_00119560.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: XBL with normal closed hold", "window_end": 68, "window_start": 8}}
|
| 48 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4755, "index": 1036, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0044_00119560_00122200.json", "segment_action": "a leader performs salsa dance: Basic step, facing the line of dance", "segment_frame_end": 79, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0044_00119560_00122200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0044_00119560_00122200.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Basic step, facing the line of dance", "window_end": 70, "window_start": 10}}
|
| 49 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4771, "index": 1052, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0049_00132200_00134760.json", "segment_action": "a leader performs salsa dance: Dile que no with normal open hold", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0049_00132200_00134760", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0049_00132200_00134760.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Dile que no with normal open hold", "window_end": 77, "window_start": 17}}
|
| 50 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4775, "index": 1056, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0051_00137280_00139840.json", "segment_action": "a leader performs salsa dance: Right turn for the follower, normal open hold, ends with Left HL for the follower", "segment_frame_end": 77, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0051_00137280_00139840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0051_00137280_00139840.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Right turn for the follower, normal open hold, ends with Left HL for the follower", "window_end": 60, "window_start": 0}}
|
| 51 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4779, "index": 1060, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair4_song1_take1__leader__seg0052_00139840_00142360.json", "segment_action": "a leader performs salsa dance: Enchufla from left HL for the follower with left turn for the leader", "segment_frame_end": 76, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair4_song1_take1__leader__seg0052_00139840_00142360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair4_song1_take1__leader__seg0052_00139840_00142360.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair4/Pair4_song1_take1 leader: Enchufla from left HL for the follower with left turn for the leader", "window_end": 68, "window_start": 8}}
|
| 52 |
+
{"caption": "", "dataset": "compas3d_original_audio_test", "global_index": 4807, "index": 1088, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/labels/Pair6_song3_take1__leader__seg0000_00000000_00150767.json", "segment_action": "a leader performs salsa dance: salsa dance", "segment_frame_end": 4523, "segment_frame_start": 0, "segment_index": 0, "segment_style": "salsa", "sequence_name": "Pair6_song3_take1__leader__seg0000_00000000_00150767", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/compas3d/g1/Pair6_song3_take1__leader__seg0000_00000000_00150767.npz", "split": "test", "text_path": null, "video_summary": "CoMPAS3D Pair6/Pair6_song3_take1 leader: salsa dance", "window_end": 4523, "window_start": 4463}}
|
| 53 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5438, "index": 21, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/007_retarget__seg0007_000840_001080.json", "segment_action": "a dancer spins, waves arms, makes hand gestures, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "007_retarget__seg0007_000840_001080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/007_retarget__seg0007_000840_001080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with spins, arm gestures, and rhythmic footwork", "window_end": 60, "window_start": 0}}
|
| 54 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5458, "index": 41, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/007_retarget__seg0013_001560_001800.json", "segment_action": "a dancer waves arms, steps side to side, spins, shifts weight, and bends knees rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "007_retarget__seg0013_001560_001800", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/007_retarget__seg0013_001560_001800.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with arm gestures, side steps, spins, and knee bends", "window_end": 240, "window_start": 180}}
|
| 55 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5475, "index": 58, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/007_retarget__seg0019_002280_002520.json", "segment_action": "a dancer raises arms, twists torso, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "007_retarget__seg0019_002280_002520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/007_retarget__seg0019_002280_002520.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with fluid arm gestures, torso twists, and rhythmic stepping", "window_end": 150, "window_start": 90}}
|
| 56 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5535, "index": 118, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/008_retarget__seg0009_001080_001320.json", "segment_action": "a dancer raises arms, sweeps them wide, spins, twists torso, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "008_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/008_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated dance with arm sweeps, spins, torso twists, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 57 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5545, "index": 128, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/008_retarget__seg0012_001440_001680.json", "segment_action": "a dancer executes varied arm gestures, twists torso, sways hips, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "008_retarget__seg0012_001440_001680", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/008_retarget__seg0012_001440_001680.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic sequence with arm gestures, torso twists, and coordinated steps", "window_end": 240, "window_start": 180}}
|
| 58 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5562, "index": 145, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/008_retarget__seg0018_002160_002400.json", "segment_action": "a dancer bends torso, sweeps arms, shifts weight, bends knees, and turns rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive fluid", "sequence_name": "008_retarget__seg0018_002160_002400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/008_retarget__seg0018_002160_002400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fluid dance with torso bends, arm sweeps, and rhythmic weight shifts", "window_end": 150, "window_start": 90}}
|
| 59 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5627, "index": 210, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/012_retarget__seg0008_000960_001200.json", "segment_action": "a dancer waves arms, twists torso, shifts weight, steps side to side, and spins", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "012_retarget__seg0008_000960_001200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/012_retarget__seg0008_000960_001200.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated dance with arm motions, torso twists, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 60 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5631, "index": 214, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/012_retarget__seg0009_001080_001320.json", "segment_action": "a dancer spins, waves arms, sways hips, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "012_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/012_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with spins, arm gestures, hip sways, and rhythmic stepping", "window_end": 150, "window_start": 90}}
|
| 61 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5650, "index": 233, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/012_retarget__seg0015_001800_002040.json", "segment_action": "a dancer bends knees, swings arms, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "012_retarget__seg0015_001800_002040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/012_retarget__seg0015_001800_002040.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic dance with knee bends, arm swings, torso twists, and rhythmic weight shifts", "window_end": 240, "window_start": 180}}
|
| 62 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5657, "index": 240, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/012_retarget__seg0018_002160_002400.json", "segment_action": "a dancer raises arms, twists torso, sways hips, bends knees, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "012_retarget__seg0018_002160_002400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/012_retarget__seg0018_002160_002400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive, coordinated dance with arm movements, torso twists, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 63 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5680, "index": 263, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0002_000240_000480.json", "segment_action": "a dancer waves arms, twists torso, steps side to side, jumps, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "014_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated dance with arm gestures, torso twists, jumps, and rhythmic steps", "window_end": 240, "window_start": 180}}
|
| 64 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5683, "index": 266, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0003_000360_000600.json", "segment_action": "a dancer gestures with arms, sways hips, steps side to side, twists torso, shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0003_000360_000600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0003_000360_000600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, hip sways, side steps, and torso twists", "window_end": 240, "window_start": 180}}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_005/samples.jsonl
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5697, "index": 280, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0008_000960_001200.json", "segment_action": "a dancer waves arms, twists torso, sways hips, steps side to side rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic coordinated", "sequence_name": "014_retarget__seg0008_000960_001200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0008_000960_001200.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, and coordinated steps", "window_end": 150, "window_start": 90}}
|
| 2 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5702, "index": 285, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0010_001200_001440.json", "segment_action": "a dancer twists torso, steps side to side, moves arms, shifts weight, and turns rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0010_001200_001440", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0010_001200_001440.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with torso twists, side steps, arm movements, and body turns", "window_end": 60, "window_start": 0}}
|
| 3 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5712, "index": 295, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0013_001560_001800.json", "segment_action": "a dancer makes arm gestures, twists torso, bends knees, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0013_001560_001800", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0013_001560_001800.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso twists, and coordinated footwork", "window_end": 150, "window_start": 90}}
|
| 4 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5715, "index": 298, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0014_001680_001920.json", "segment_action": "a dancer waves arms, twists torso, bends knees, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0014_001680_001920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0014_001680_001920.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso twists, and coordinated footwork", "window_end": 150, "window_start": 90}}
|
| 5 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5718, "index": 301, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0015_001800_002040.json", "segment_action": "a dancer makes arm gestures, twists torso, bends knees, and steps side to side", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0015_001800_002040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0015_001800_002040.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, torso twists, knee bends, and side steps", "window_end": 150, "window_start": 90}}
|
| 6 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5720, "index": 303, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0016_001920_002160.json", "segment_action": "a dancer bends knees, sways hips, swings arms, steps side to side, twists torso rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0016_001920_002160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0016_001920_002160.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with knee bends, arm swings, hip sways, and side steps", "window_end": 60, "window_start": 0}}
|
| 7 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5738, "index": 321, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0022_002640_002880.json", "segment_action": "a dancer sways hips, flicks arms, shifts weight, and turns body rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0022_002640_002880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0022_002640_002880.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with hip sways, arm flicks, and body turns", "window_end": 60, "window_start": 0}}
|
| 8 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5750, "index": 333, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0026_003120_003360.json", "segment_action": "a dancer steps, twists torso, raises arms, spins, and shifts weight with rhythmic knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "014_retarget__seg0026_003120_003360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0026_003120_003360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with rhythmic steps, torso twists, arm gestures, and spins", "window_end": 60, "window_start": 0}}
|
| 9 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5756, "index": 339, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0028_003360_003600.json", "segment_action": "a dancer swings arms, twists torso, steps side to side, and shifts weight with coordinated movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0028_003360_003600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0028_003360_003600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, and coordinated stepping", "window_end": 60, "window_start": 0}}
|
| 10 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5758, "index": 341, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0028_003360_003600.json", "segment_action": "a dancer swings arms, twists torso, steps side to side, and shifts weight with coordinated movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "014_retarget__seg0028_003360_003600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0028_003360_003600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, and coordinated stepping", "window_end": 240, "window_start": 180}}
|
| 11 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5767, "index": 350, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0031_003720_003960.json", "segment_action": "a dancer sways hips, waves arms, steps rhythmically, twists torso, and executes a spin", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "014_retarget__seg0031_003720_003960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0031_003720_003960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated dance with hip sways, arm gestures, rhythmic steps, and a spin", "window_end": 240, "window_start": 180}}
|
| 12 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5801, "index": 384, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0043_005160_005400.json", "segment_action": "a dancer spins, bends knees, twists torso, and moves arms with rhythmic weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "014_retarget__seg0043_005160_005400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0043_005160_005400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated sequence with spins, knee bends, torso twists, and arm gestures", "window_end": 60, "window_start": 0}}
|
| 13 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5804, "index": 387, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/014_retarget__seg0044_005280_005520.json", "segment_action": "a dancer bends knees, sweeps arms, twists torso, shifts weight, and executes a full turn", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "014_retarget__seg0044_005280_005520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/014_retarget__seg0044_005280_005520.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated dance with knee bends, arm gestures, weight shifts, and a turn", "window_end": 60, "window_start": 0}}
|
| 14 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5840, "index": 423, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0000_000000_000240.json", "segment_action": "a dancer sways hips, moves arms, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful coordinated", "sequence_name": "019_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful coordinated sequence with hip sways, arm movements, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 15 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5846, "index": 429, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0002_000240_000480.json", "segment_action": "a dancer turns their torso, steps side to side, and shifts weight with controlled limb movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "smooth coordinated", "sequence_name": "019_retarget__seg0002_000240_000480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0002_000240_000480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a smooth coordinated sequence with body turns, side steps, and rhythmic weight shifts", "window_end": 60, "window_start": 0}}
|
| 16 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5884, "index": 467, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0014_001680_001920.json", "segment_action": "a dancer kneels, moves arms to head, twists torso, and turns head rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive graceful", "sequence_name": "019_retarget__seg0014_001680_001920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0014_001680_001920.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fine dance sequence with arm gestures, torso twists, and head turns while kneeling", "window_end": 240, "window_start": 180}}
|
| 17 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5888, "index": 471, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0017_002040_002280.json", "segment_action": "a dancer kneels, moves arms dynamically, twists torso, then stands and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "019_retarget__seg0017_002040_002280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0017_002040_002280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic dance with arm movements, torso twists, and a kneel-to-stand transition", "window_end": 60, "window_start": 0}}
|
| 18 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5906, "index": 489, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0023_002760_003000.json", "segment_action": "a dancer sways hips, makes fluid arm gestures, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "019_retarget__seg0023_002760_003000", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0023_002760_003000.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with hip sways, arm gestures, and side steps", "window_end": 60, "window_start": 0}}
|
| 19 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5939, "index": 522, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0035_004200_004440.json", "segment_action": "a dancer steps forward, raises arms, twists torso, and does fluid arm gestures with leg coordination", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "019_retarget__seg0035_004200_004440", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0035_004200_004440.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive fine dance sequence with fluid arm movements and coordinated steps", "window_end": 60, "window_start": 0}}
|
| 20 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5952, "index": 535, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0039_004680_004920.json", "segment_action": "A dancer kneels, rises with flowing arm gestures, steps, turns, and walks forward rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive graceful", "sequence_name": "019_retarget__seg0039_004680_004920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0039_004680_004920.npz", "split": "test", "text_path": null, "video_summary": "A dancer performs an expressive graceful sequence with kneeling, rising, arm movements, steps, turns, and walking", "window_end": 150, "window_start": 90}}
|
| 21 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5955, "index": 538, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0040_004800_005040.json", "segment_action": "a dancer raises arms, steps, twists torso, turns, and does fluid arm gestures", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "019_retarget__seg0040_004800_005040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0040_004800_005040.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful fine dance sequence with arm movements, steps, torso twists, and turns", "window_end": 150, "window_start": 90}}
|
| 22 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5964, "index": 547, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0043_005160_005400.json", "segment_action": "a dancer steps side to side, swings arms, twists torso, sways hips, and bends knees rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "019_retarget__seg0043_005160_005400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0043_005160_005400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with coordinated steps, arm swings, torso twists, and hip sways", "window_end": 150, "window_start": 90}}
|
| 23 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5967, "index": 550, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0046_005520_005760.json", "segment_action": "a dancer transitions from kneeling to standing, swaying hips and moving arms with rhythmic steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "019_retarget__seg0046_005520_005760", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0046_005520_005760.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic sequence transitioning from kneeling to standing with hip sways and arm movements", "window_end": 150, "window_start": 90}}
|
| 24 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5974, "index": 557, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/019_retarget__seg0048_005760_006000.json", "segment_action": "a dancer swings arms, sways hips, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "019_retarget__seg0048_005760_006000", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/019_retarget__seg0048_005760_006000.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated dance with arm swings, hip sways, and rhythmic stepping", "window_end": 240, "window_start": 180}}
|
| 25 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 5980, "index": 563, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/033_retarget__seg0001_000120_000360.json", "segment_action": "a dancer waves arms, twists torso, sways hips, turns body, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy expressive rhythmic", "sequence_name": "033_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/033_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy, expressive dance with arm gestures, torso twists, and body turns", "window_end": 240, "window_start": 180}}
|
| 26 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6002, "index": 585, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/033_retarget__seg0009_001080_001320.json", "segment_action": "a dancer moves arms fluidly, twists torso, sways hips, and steps with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "033_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/033_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive dance with fluid arm movements, torso twists, and coordinated footwork", "window_end": 60, "window_start": 0}}
|
| 27 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6036, "index": 619, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/033_retarget__seg0025_003000_003240.json", "segment_action": "a dancer moves arms fluidly, twists torso, steps, spins, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "033_retarget__seg0025_003000_003240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/033_retarget__seg0025_003000_003240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive dance with fluid arm gestures, torso twists, steps, and spins", "window_end": 150, "window_start": 90}}
|
| 28 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6042, "index": 625, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/033_retarget__seg0027_003240_003480.json", "segment_action": "a dancer spins, extends arms in varied gestures, steps rhythmically, twists torso", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "033_retarget__seg0027_003240_003480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/033_retarget__seg0027_003240_003480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with spins, arm gestures, and rhythmic stepping", "window_end": 150, "window_start": 90}}
|
| 29 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6045, "index": 628, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/040_retarget__seg0000_000000_000240.json", "segment_action": "a dancer does fluid arm waves, torso sways, and gentle weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "smooth graceful", "sequence_name": "040_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/040_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful smooth sequence with fluid arm waves and body sways", "window_end": 150, "window_start": 90}}
|
| 30 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6072, "index": 655, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/040_retarget__seg0009_001080_001320.json", "segment_action": "a dancer spins, sweeps arms, kicks legs, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "040_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/040_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with spins, arm sweeps, leg kicks, and weight shifts", "window_end": 150, "window_start": 90}}
|
| 31 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6105, "index": 688, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/051_retarget__seg0012_001440_001680.json", "segment_action": "a dancer executes fluid arm movements, torso bends, weight shifts, and precise footwork", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "051_retarget__seg0012_001440_001680", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/051_retarget__seg0012_001440_001680.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive fine dance with fluid arm gestures, torso bends, and controlled footwork", "window_end": 150, "window_start": 90}}
|
| 32 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6122, "index": 705, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/051_retarget__seg0018_002160_002400.json", "segment_action": "a dancer bends knees, kicks a leg, swings arms, twists torso, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dynamic", "sequence_name": "051_retarget__seg0018_002160_002400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/051_retarget__seg0018_002160_002400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dynamic fine dance with leg kicks, arm swings, torso twists, and rhythmic footwork", "window_end": 60, "window_start": 0}}
|
| 33 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6140, "index": 723, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/051_retarget__seg0025_003000_003240.json", "segment_action": "a dancer bends knees, raises arms, kicks legs, twists torso, and dips forward", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic expressive", "sequence_name": "051_retarget__seg0025_003000_003240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/051_retarget__seg0025_003000_003240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic expressive sequence with knee bends, arm gestures, leg kicks, and torso twists", "window_end": 60, "window_start": 0}}
|
| 34 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6143, "index": 726, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/051_retarget__seg0026_003120_003360.json", "segment_action": "a dancer executes high kicks, sweeps arms, bends torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dynamic playful", "sequence_name": "051_retarget__seg0026_003120_003360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/051_retarget__seg0026_003120_003360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a dynamic playful dance with high kicks, arm sweeps, and torso bends", "window_end": 60, "window_start": 0}}
|
| 35 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6153, "index": 736, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/051_retarget__seg0029_003480_003720.json", "segment_action": "a dancer sweeps arms fluidly, twists torso, and shifts weight with rhythmic steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "051_retarget__seg0029_003480_003720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/051_retarget__seg0029_003480_003720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive dance with fluid arm sweeps, torso twists, and rhythmic weight shifts", "window_end": 150, "window_start": 90}}
|
| 36 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6237, "index": 820, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/071_retarget__seg0017_002040_002280.json", "segment_action": "a dancer uses varied arm gestures, twists torso, bends knees, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "071_retarget__seg0017_002040_002280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/071_retarget__seg0017_002040_002280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, and dynamic footwork", "window_end": 150, "window_start": 90}}
|
| 37 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6286, "index": 869, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/071_retarget__seg0033_003960_004200.json", "segment_action": "a dancer bends knees, waves arms, twists torso, shifts weight, and turns directionally", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "071_retarget__seg0033_003960_004200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/071_retarget__seg0033_003960_004200.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with knee bends, arm gestures, and directional turns", "window_end": 240, "window_start": 180}}
|
| 38 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6295, "index": 878, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/071_retarget__seg0036_004320_004560.json", "segment_action": "a dancer bends knees, twists torso, waves arms, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "071_retarget__seg0036_004320_004560", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/071_retarget__seg0036_004320_004560.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated dance with knee bends, torso twists, and arm gestures", "window_end": 240, "window_start": 180}}
|
| 39 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6301, "index": 884, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/071_retarget__seg0038_004560_004800.json", "segment_action": "a dancer twists torso, sways hips, makes varied arm gestures, and shifts weight with rhythmic steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "071_retarget__seg0038_004560_004800", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/071_retarget__seg0038_004560_004800.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with torso twists, arm gestures, and rhythmic footwork", "window_end": 240, "window_start": 180}}
|
| 40 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6370, "index": 953, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0015_001800_002040.json", "segment_action": "a dancer waves arms, sways hips, steps side to side, and twists torso with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0015_001800_002040", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0015_001800_002040.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm waves, hip sways, and coordinated steps", "window_end": 240, "window_start": 180}}
|
| 41 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6371, "index": 954, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0016_001920_002160.json", "segment_action": "a dancer moves arms, twists torso, sways hips, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "100_retarget__seg0016_001920_002160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0016_001920_002160.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive, coordinated dance with arm gestures, torso twists, and dynamic steps", "window_end": 60, "window_start": 0}}
|
| 42 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6377, "index": 960, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0018_002160_002400.json", "segment_action": "a dancer makes varied arm gestures, twists torso, sways hips, and shifts weight with knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0018_002160_002400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0018_002160_002400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, and knee-bent weight shifts", "window_end": 60, "window_start": 0}}
|
| 43 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6381, "index": 964, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0019_002280_002520.json", "segment_action": "a dancer swings arms, sways hips, steps side to side, and twists torso with rhythmic shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0019_002280_002520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0019_002280_002520.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm swings, hip sways, and side steps", "window_end": 150, "window_start": 90}}
|
| 44 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6404, "index": 987, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0027_003240_003480.json", "segment_action": "a dancer moves arms fluidly, sways hips, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0027_003240_003480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0027_003240_003480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with fluid arm movements, hip sways, and rhythmic stepping", "window_end": 60, "window_start": 0}}
|
| 45 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6413, "index": 996, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0030_003600_003840.json", "segment_action": "a dancer makes fluid arm gestures, sways hips, and steps side to side with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0030_003600_003840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0030_003600_003840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with fluid arm gestures, hip sways, and side steps", "window_end": 60, "window_start": 0}}
|
| 46 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6419, "index": 1002, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0032_003840_004080.json", "segment_action": "a dancer kneels, arches torso, waves arms, twists body, and shifts weight on knees", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive dramatic", "sequence_name": "100_retarget__seg0032_003840_004080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0032_003840_004080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive dramatic sequence with kneeling, torso arches, arm gestures, and body twists", "window_end": 60, "window_start": 0}}
|
| 47 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6449, "index": 1032, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0042_005040_005280.json", "segment_action": "a dancer executes arm crosses, hip sways, knee bends, and rhythmic foot movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "100_retarget__seg0042_005040_005280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0042_005040_005280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic sequence with arm gestures, hip sways, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 48 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6450, "index": 1033, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0042_005040_005280.json", "segment_action": "a dancer executes arm crosses, hip sways, knee bends, and rhythmic foot movements", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "100_retarget__seg0042_005040_005280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0042_005040_005280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic sequence with arm gestures, hip sways, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 49 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6452, "index": 1035, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0043_005160_005400.json", "segment_action": "a dancer sways hips, moves arms in fluid gestures, steps side to side, and twists torso", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0043_005160_005400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0043_005160_005400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with hip sways, arm gestures, and coordinated footwork", "window_end": 60, "window_start": 0}}
|
| 50 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6480, "index": 1063, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0052_006240_006480.json", "segment_action": "a dancer moves arms, twists torso, sways hips, steps side to side with bent knees", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "100_retarget__seg0052_006240_006480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0052_006240_006480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, hip sways, and stepping", "window_end": 150, "window_start": 90}}
|
| 51 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6482, "index": 1065, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0053_006360_006600.json", "segment_action": "a dancer steps side to side, waves arms, lifts knees, twists torso, shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "rhythmic groovy", "sequence_name": "100_retarget__seg0053_006360_006600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0053_006360_006600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a rhythmic groovy sequence with side steps, arm gestures, knee lifts, and body turns", "window_end": 60, "window_start": 0}}
|
| 52 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6485, "index": 1068, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/100_retarget__seg0054_006480_006720.json", "segment_action": "a dancer steps rhythmically, bends knees, swings arms, and shifts weight with turns", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic energetic", "sequence_name": "100_retarget__seg0054_006480_006720", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/100_retarget__seg0054_006480_006720.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an energetic groovy dance with rhythmic steps, arm swings, and knee bends", "window_end": 60, "window_start": 0}}
|
| 53 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6500, "index": 1083, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0000_000000_000240.json", "segment_action": "a dancer fluidly moves arms, twists torso, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful coordinated", "sequence_name": "108_retarget__seg0000_000000_000240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0000_000000_000240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful coordinated sequence with fluid arm movements, torso twists, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 54 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6517, "index": 1100, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0005_000600_000840.json", "segment_action": "a dancer kneels, raises arms, stands, twists torso, and does fluid arm extensions and steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "108_retarget__seg0005_000600_000840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0005_000600_000840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with fluid arm movements, torso twists, and kneel-to-stand transitions", "window_end": 240, "window_start": 180}}
|
| 55 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6521, "index": 1104, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0007_000840_001080.json", "segment_action": "a dancer waves arms, bends torso, sways hips, and takes rhythmic side steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive graceful", "sequence_name": "108_retarget__seg0007_000840_001080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0007_000840_001080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive graceful sequence with arm waves, torso bends, hip sways, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 56 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6524, "index": 1107, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0008_000960_001200.json", "segment_action": "a dancer sweeps arms fluidly, twists torso, shifts weight, and holds controlled leg stances", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "108_retarget__seg0008_000960_001200", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0008_000960_001200.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive fine dance with fluid arm movements, torso twists, and weight shifts", "window_end": 60, "window_start": 0}}
|
| 57 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6539, "index": 1122, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0014_001680_001920.json", "segment_action": "a dancer executes fluid arm gestures, sways torso, shifts weight, and steps with gentle turns", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "108_retarget__seg0014_001680_001920", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0014_001680_001920.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive fine dance with fluid arm movements, torso sways, and coordinated steps", "window_end": 60, "window_start": 0}}
|
| 58 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6567, "index": 1150, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0023_002760_003000.json", "segment_action": "a dancer raises arms, bends knees, squats, twists torso, and shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive graceful", "sequence_name": "108_retarget__seg0023_002760_003000", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0023_002760_003000.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fine dance with arm gestures, knee bends, squats, and body twists", "window_end": 150, "window_start": 90}}
|
| 59 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6568, "index": 1151, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/108_retarget__seg0023_002760_003000.json", "segment_action": "a dancer raises arms, bends knees, squats, twists torso, and shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive graceful", "sequence_name": "108_retarget__seg0023_002760_003000", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/108_retarget__seg0023_002760_003000.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fine dance with arm gestures, knee bends, squats, and body twists", "window_end": 240, "window_start": 180}}
|
| 60 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6599, "index": 1182, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0009_001080_001320.json", "segment_action": "a dancer raises arms, sways hips, bends knees, and takes rhythmic side steps with weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "113_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with arm raises, hip sways, knee bends, and side steps", "window_end": 60, "window_start": 0}}
|
| 61 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6632, "index": 1215, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0020_002400_002640.json", "segment_action": "a dancer moves arms fluidly, twists torso, steps side to side, and spins with coordination", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0020_002400_002640", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0020_002400_002640.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive fine dance with arm gestures, torso twists, steps, and spins", "window_end": 60, "window_start": 0}}
|
| 62 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6638, "index": 1221, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0022_002640_002880.json", "segment_action": "a dancer moves arms fluidly, twists torso, sways hips, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0022_002640_002880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0022_002640_002880.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with fluid arm gestures, torso twists, and rhythmic stepping", "window_end": 60, "window_start": 0}}
|
| 63 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6639, "index": 1222, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0022_002640_002880.json", "segment_action": "a dancer moves arms fluidly, twists torso, sways hips, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0022_002640_002880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0022_002640_002880.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with fluid arm gestures, torso twists, and rhythmic stepping", "window_end": 150, "window_start": 90}}
|
| 64 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6640, "index": 1223, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0022_002640_002880.json", "segment_action": "a dancer moves arms fluidly, twists torso, sways hips, and steps rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0022_002640_002880", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0022_002640_002880.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive sequence with fluid arm gestures, torso twists, and rhythmic stepping", "window_end": 240, "window_start": 180}}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_006/samples.jsonl
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6652, "index": 1235, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0030_003600_003840.json", "segment_action": "a dancer spins, sweeps arms, shifts weight, and steps rhythmically with body turns", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful coordinated", "sequence_name": "113_retarget__seg0030_003600_003840", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0030_003600_003840.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful coordinated dance with spins, arm sweeps, and rhythmic stepping", "window_end": 240, "window_start": 180}}
|
| 2 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6660, "index": 1243, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0035_004200_004440.json", "segment_action": "a dancer steps, spins, waves arms, twists torso, and shifts weight with hip sways", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "113_retarget__seg0035_004200_004440", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0035_004200_004440.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with steps, spins, arm gestures, and hip sways", "window_end": 150, "window_start": 90}}
|
| 3 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6678, "index": 1261, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0041_004920_005160.json", "segment_action": "a dancer waves arms, bends torso, steps rhythmically, spins, and coordinates limbs", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive fluid", "sequence_name": "113_retarget__seg0041_004920_005160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0041_004920_005160.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fluid dance with arm gestures, torso bends, steps, and spins", "window_end": 150, "window_start": 90}}
|
| 4 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6679, "index": 1262, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0041_004920_005160.json", "segment_action": "a dancer waves arms, bends torso, steps rhythmically, spins, and coordinates limbs", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive fluid", "sequence_name": "113_retarget__seg0041_004920_005160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0041_004920_005160.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive fluid dance with arm gestures, torso bends, steps, and spins", "window_end": 240, "window_start": 180}}
|
| 5 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6706, "index": 1289, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0050_006000_006240.json", "segment_action": "a dancer executes fluid arm gestures, torso twists, side steps, and coordinated knee bends", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0050_006000_006240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0050_006000_006240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful, expressive dance with fluid arm movements, torso twists, and rhythmic steps", "window_end": 240, "window_start": 180}}
|
| 6 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6714, "index": 1297, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0053_006360_006600.json", "segment_action": "a dancer spins, raises arms, waves hands, bends knees, and steps rhythmically with torso twists", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful coordinated", "sequence_name": "113_retarget__seg0053_006360_006600", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0053_006360_006600.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful coordinated fine dance with spins, arm gestures, and rhythmic footwork", "window_end": 150, "window_start": 90}}
|
| 7 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6738, "index": 1321, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/113_retarget__seg0061_007320_007560.json", "segment_action": "a dancer twists torso, bends knees, moves arms, and holds a stretched lean pose", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful expressive", "sequence_name": "113_retarget__seg0061_007320_007560", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/113_retarget__seg0061_007320_007560.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful expressive dance with torso twists, arm gestures, and a stretched pose", "window_end": 150, "window_start": 90}}
|
| 8 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6743, "index": 1326, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0001_000120_000360.json", "segment_action": "a dancer sways hips, uses varied arm gestures, steps side to side, and bends forward", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "143_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with arm gestures, hip sways, side steps, and a forward bend", "window_end": 60, "window_start": 0}}
|
| 9 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6759, "index": 1342, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0006_000720_000960.json", "segment_action": "a dancer waves arms, twists torso, takes side steps, turns, and shifts weight smoothly", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful coordinated", "sequence_name": "143_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful fine dance with arm gestures, torso twists, coordinated steps and turns", "window_end": 150, "window_start": 90}}
|
| 10 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6767, "index": 1350, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0009_001080_001320.json", "segment_action": "a dancer waves one arm, twists torso, bends knees, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive smooth", "sequence_name": "143_retarget__seg0009_001080_001320", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0009_001080_001320.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive smooth dance with arm waves, torso twists, and knee bends", "window_end": 60, "window_start": 0}}
|
| 11 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6771, "index": 1354, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0010_001200_001440.json", "segment_action": "a dancer raises arms overhead, bends torso sideways, lifts one knee, then lowers arms", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful controlled", "sequence_name": "143_retarget__seg0010_001200_001440", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0010_001200_001440.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful sequence with arm raises, torso bends, and a knee lift", "window_end": 150, "window_start": 90}}
|
| 12 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6776, "index": 1359, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0012_001440_001680.json", "segment_action": "a dancer waves arms, sways hips, steps, turns, and moves with coordinated motions", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful rhythmic", "sequence_name": "143_retarget__seg0012_001440_001680", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0012_001440_001680.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful rhythmic dance with arm gestures, hip sways, and stepping turns", "window_end": 60, "window_start": 0}}
|
| 13 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6874, "index": 1457, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/143_retarget__seg0044_005280_005520.json", "segment_action": "a dancer spins 360 degrees, holds one arm overhead, steps forward, then lowers the arm", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "graceful controlled", "sequence_name": "143_retarget__seg0044_005280_005520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/143_retarget__seg0044_005280_005520.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a graceful, controlled sequence with a full spin, raised arm, and forward steps", "window_end": 240, "window_start": 180}}
|
| 14 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6937, "index": 1520, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/162_retarget__seg0018_002160_002400.json", "segment_action": "a dancer bends knees, steps side to side, twists torso, moves arms, and spins", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy coordinated", "sequence_name": "162_retarget__seg0018_002160_002400", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/162_retarget__seg0018_002160_002400.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy coordinated sequence with knee bends, side steps, arm gestures, and spins", "window_end": 240, "window_start": 180}}
|
| 15 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6940, "index": 1523, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/162_retarget__seg0019_002280_002520.json", "segment_action": "a dancer spins, waves arms, bends knees, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "162_retarget__seg0019_002280_002520", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/162_retarget__seg0019_002280_002520.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with spins, arm gestures, knee bends, and weight shifts", "window_end": 240, "window_start": 180}}
|
| 16 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 6996, "index": 1579, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/178_retarget__seg0001_000120_000360.json", "segment_action": "a dancer gestures with arms, twists torso, sways hips, steps side to side, shifts weight", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "178_retarget__seg0001_000120_000360", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/178_retarget__seg0001_000120_000360.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic dance with arm gestures, torso twists, hip sways, and coordinated steps", "window_end": 150, "window_start": 90}}
|
| 17 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7010, "index": 1593, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/178_retarget__seg0006_000720_000960.json", "segment_action": "a dancer moves arms, sways hips, steps side to side, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "178_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/178_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm gestures, hip sways, and side steps", "window_end": 60, "window_start": 0}}
|
| 18 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7031, "index": 1614, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0006_000720_000960.json", "segment_action": "a dancer raises arms, bends torso, spins, and shifts weight with rhythmic stepping", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "184_retarget__seg0006_000720_000960", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0006_000720_000960.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic dance with arm raises, torso bends, spins, and coordinated footwork", "window_end": 60, "window_start": 0}}
|
| 19 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7034, "index": 1617, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0007_000840_001080.json", "segment_action": "a dancer bends torso, spins, waves arms, steps side to side, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated rhythmic", "sequence_name": "184_retarget__seg0007_000840_001080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0007_000840_001080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with torso bends, spins, arm gestures, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 20 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7066, "index": 1649, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0017_002040_002280.json", "segment_action": "a dancer sways hips, gestures with hands, steps side to side, and spins with extended arms", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful coordinated", "sequence_name": "184_retarget__seg0017_002040_002280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0017_002040_002280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful coordinated dance with hip sways, hand gestures, side steps, and spins", "window_end": 240, "window_start": 180}}
|
| 21 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7079, "index": 1662, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0025_003000_003240.json", "segment_action": "a dancer waves arms, gestures hands, bends knees, twists torso, and shifts weight rhythmically", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "playful expressive", "sequence_name": "184_retarget__seg0025_003000_003240", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0025_003000_003240.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a playful expressive dance with dynamic arm gestures, knee bends, and rhythmic steps", "window_end": 60, "window_start": 0}}
|
| 22 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7085, "index": 1668, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0027_003240_003480.json", "segment_action": "a dancer steps side to side, bends torso, waves arms, and makes dynamic hand gestures", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "184_retarget__seg0027_003240_003480", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0027_003240_003480.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic sequence with side steps, torso bends, and dynamic arm movements", "window_end": 60, "window_start": 0}}
|
| 23 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7101, "index": 1684, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0032_003840_004080.json", "segment_action": "a dancer executes arm waves, torso twists, spins, knee bends, and rhythmic weight shifts", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "184_retarget__seg0032_003840_004080", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0032_003840_004080.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated dance with arm gestures, spins, and rhythmic steps", "window_end": 150, "window_start": 90}}
|
| 24 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7111, "index": 1694, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/184_retarget__seg0035_004200_004440.json", "segment_action": "a dancer twists torso, waves arms, steps side to side, and shifts weight while turning", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "groovy rhythmic", "sequence_name": "184_retarget__seg0035_004200_004440", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/184_retarget__seg0035_004200_004440.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs a groovy rhythmic sequence with arm movements, torso twists, side steps, and body turns", "window_end": 240, "window_start": 180}}
|
| 25 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7152, "index": 1735, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/211_retarget__seg0016_001920_002160.json", "segment_action": "a dancer spins, steps side to side, squats low, and makes coordinated arm gestures", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive rhythmic", "sequence_name": "211_retarget__seg0016_001920_002160", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/211_retarget__seg0016_001920_002160.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive rhythmic sequence with spins, steps, squats, and arm gestures", "window_end": 150, "window_start": 90}}
|
| 26 |
+
{"caption": "", "dataset": "finedance_original_audio_test", "global_index": 7155, "index": 1738, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/labels/211_retarget__seg0017_002040_002280.json", "segment_action": "a dancer kneels, makes sweeping arm gestures, shifts torso, then stands and steps", "segment_frame_end": 240, "segment_frame_start": 0, "segment_index": 0, "segment_style": "expressive coordinated", "sequence_name": "211_retarget__seg0017_002040_002280", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/finedance/g1/211_retarget__seg0017_002040_002280.npz", "split": "test", "text_path": null, "video_summary": "a dancer performs an expressive coordinated fine dance with kneeling, arm movements, torso shifts, and standing steps", "window_end": 150, "window_start": 90}}
|
| 27 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4834, "index": 14, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__00381_slice18.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent-knee stances with weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__00381_slice18", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__00381_slice18.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent-knee stances with weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 300, "window_start": 240}}
|
| 28 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4835, "index": 15, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__00398_slice9.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with syncopated steps; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__00398_slice9", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__00398_slice9.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with syncopated steps; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 60, "window_start": 0}}
|
| 29 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4858, "index": 38, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01114_slice13.json", "segment_action": "The dancer is performing a National modern dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with torsional spine motions; #head_movement: Sharp tilts and rotations aligned with directional changes; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Quick weight shifts and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01114_slice13", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01114_slice13.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National modern dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with torsional spine motions; #head_movement: Sharp tilts and rotations aligned with directional changes; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Quick weight shifts and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 300, "window_start": 240}}
|
| 30 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4869, "index": 49, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01315_slice1.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick tilts and turns in sync with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Swift weight transfers and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01315_slice1", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01315_slice1.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick tilts and turns in sync with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Swift weight transfers and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "window_end": 180, "window_start": 120}}
|
| 31 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4877, "index": 57, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01534_slice13.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01534_slice13", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01534_slice13.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 60, "window_start": 0}}
|
| 32 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4878, "index": 58, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01534_slice13.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01534_slice13", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01534_slice13.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 180, "window_start": 120}}
|
| 33 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4893, "index": 73, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01725_slice19.json", "segment_action": "The dancer is performing a Street dance with Jazz style. #movement_style: Fast and complicated; #emotional_expression: Dynamic and sharp; #whole_body_movement: Strong isolations and quick directional changes; #head_movement: Precise tilts and pops in sync with beats; #arms_movement: Extended, fluid gestures with sudden staccato motions; #legs_movement: Rapid steps and controlled weight shifts; #camera_angle: Frontal view; #camera_distance: Full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01725_slice19", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01725_slice19.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a Street dance with Jazz style. #movement_style: Fast and complicated; #emotional_expression: Dynamic and sharp; #whole_body_movement: Strong isolations and quick directional changes; #head_movement: Precise tilts and pops in sync with beats; #arms_movement: Extended, fluid gestures with sudden staccato motions; #legs_movement: Rapid steps and controlled weight shifts; #camera_angle: Frontal view; #camera_distance: Full-body visible.", "window_end": 180, "window_start": 120}}
|
| 34 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4906, "index": 86, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01859_slice13.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid yet rhythmic isolations with strong downbeats; #emotional_expression: confident and grounded; #whole_body_movement: dynamic vertical bounce with controlled sways; #head_movement: sharp nods in sync with beats; #arms_movement: alternating flexed extensions and close-to-body gestures; #legs_movement: bent-knee steps with subtle weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01859_slice13", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01859_slice13.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid yet rhythmic isolations with strong downbeats; #emotional_expression: confident and grounded; #whole_body_movement: dynamic vertical bounce with controlled sways; #head_movement: sharp nods in sync with beats; #arms_movement: alternating flexed extensions and close-to-body gestures; #legs_movement: bent-knee steps with subtle weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 300, "window_start": 240}}
|
| 35 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4909, "index": 89, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__01876_slice13.json", "segment_action": "The dancer is performing a Street dance with Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with frequent weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__01876_slice13", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__01876_slice13.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a Street dance with Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with frequent weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 300, "window_start": 240}}
|
| 36 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4910, "index": 90, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02010_slice0.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp isolations; #head_movement: Quick directional changes synchronized with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02010_slice0", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02010_slice0.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp isolations; #head_movement: Quick directional changes synchronized with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Precise weight shifts and agile steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "window_end": 60, "window_start": 0}}
|
| 37 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4914, "index": 94, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02083_slice7.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Controlled tilts synchronized with torso shifts; #arms_movement: Rapid extensions and precise articulations; #legs_movement: Quick steps and weight shifts with strong grounding; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02083_slice7", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02083_slice7.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Controlled tilts synchronized with torso shifts; #arms_movement: Rapid extensions and precise articulations; #legs_movement: Quick steps and weight shifts with strong grounding; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 180, "window_start": 120}}
|
| 38 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4918, "index": 98, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02106_slice8.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Extended, precise gestures with rapid repositioning; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02106_slice8", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02106_slice8.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick directional changes aligned with torso; #arms_movement: Extended, precise gestures with rapid repositioning; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 300, "window_start": 240}}
|
| 39 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4919, "index": 99, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02109_slice0.json", "segment_action": "The dancer is performing a National modern dance with modern style. movement_style: fast and complicated; emotional_expression: intense and focused; whole_body_movement: dynamic shifts with sharp transitions; head_movement: aligned with torso, minimal independent motion; arms_movement: extended and precise with rapid repositioning; legs_movement: quick steps and high lifts; camera_angle: front view; camera_distance: full body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02109_slice0", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02109_slice0.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National modern dance with modern style. movement_style: fast and complicated; emotional_expression: intense and focused; whole_body_movement: dynamic shifts with sharp transitions; head_movement: aligned with torso, minimal independent motion; arms_movement: extended and precise with rapid repositioning; legs_movement: quick steps and high lifts; camera_angle: front view; camera_distance: full body visible.", "window_end": 60, "window_start": 0}}
|
| 40 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4930, "index": 110, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02179_slice15.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with torsional spine motions; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Swift weight transfers and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02179_slice15", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02179_slice15.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with torsional spine motions; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and contractions; #legs_movement: Swift weight transfers and precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 300, "window_start": 240}}
|
| 41 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4931, "index": 111, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02210_slice2.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Aligned with torso, minimal independent motion; #arms_movement: Extended and precise, frequent asymmetrical positioning; #legs_movement: Rapid weight transfers, high knee lifts and quick steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02210_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02210_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Aligned with torso, minimal independent motion; #arms_movement: Extended and precise, frequent asymmetrical positioning; #legs_movement: Rapid weight transfers, high knee lifts and quick steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "window_end": 60, "window_start": 0}}
|
| 42 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4933, "index": 113, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02210_slice2.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Aligned with torso, minimal independent motion; #arms_movement: Extended and precise, frequent asymmetrical positioning; #legs_movement: Rapid weight transfers, high knee lifts and quick steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02210_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02210_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Aligned with torso, minimal independent motion; #arms_movement: Extended and precise, frequent asymmetrical positioning; #legs_movement: Rapid weight transfers, high knee lifts and quick steps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "window_end": 300, "window_start": 240}}
|
| 43 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4938, "index": 118, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02280_slice11.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid and grounded; #emotional_expression: confident and controlled; #whole_body_movement: dynamic torso isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and gestural with rhythmic swings; #legs_movement: bent knees with syncopated steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02280_slice11", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02280_slice11.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid and grounded; #emotional_expression: confident and controlled; #whole_body_movement: dynamic torso isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and gestural with rhythmic swings; #legs_movement: bent knees with syncopated steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 180, "window_start": 120}}
|
| 44 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4947, "index": 127, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02450_slice22.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with torsional spine motions; #head_movement: Quick directional changes aligned with upper body shifts; #arms_movement: Sharp, extended gestures with synchronized symmetry; #legs_movement: Rapid weight shifts and precise footwork; #camera_angle: Frontal view; #camera_distance: Medium shot capturing full body.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02450_slice22", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02450_slice22.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with torsional spine motions; #head_movement: Quick directional changes aligned with upper body shifts; #arms_movement: Sharp, extended gestures with synchronized symmetry; #legs_movement: Rapid weight shifts and precise footwork; #camera_angle: Frontal view; #camera_distance: Medium shot capturing full body.", "window_end": 180, "window_start": 120}}
|
| 45 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4960, "index": 140, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02619_slice3.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and precise articulations; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02619_slice3", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02619_slice3.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick directional changes aligned with torso; #arms_movement: Rapid extensions and precise articulations; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 300, "window_start": 240}}
|
| 46 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4962, "index": 142, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02623_slice2.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Jazz style. #movement_style: Fast and complicated; #emotional_expression: Confident and sharp; #whole_body_movement: Dynamic isolations with quick directional changes; #head_movement: Precise tilts and pops in sync with beats; #arms_movement: Angular and controlled gestures, often asymmetrical; #legs_movement: Rapid steps with bent knees and low stances; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02623_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02623_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Jazz style. #movement_style: Fast and complicated; #emotional_expression: Confident and sharp; #whole_body_movement: Dynamic isolations with quick directional changes; #head_movement: Precise tilts and pops in sync with beats; #arms_movement: Angular and controlled gestures, often asymmetrical; #legs_movement: Rapid steps with bent knees and low stances; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "window_end": 180, "window_start": 120}}
|
| 47 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4969, "index": 149, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02677_slice2.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick tilts and rotations synchronized with torso shifts; #arms_movement: Extended, fluid gestures with sudden stops; #legs_movement: Rapid steps and high extensions; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02677_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02677_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp accents; #head_movement: Quick tilts and rotations synchronized with torso shifts; #arms_movement: Extended, fluid gestures with sudden stops; #legs_movement: Rapid steps and high extensions; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "window_end": 300, "window_start": 240}}
|
| 48 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4978, "index": 158, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02697_slice17.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick tilts and rotations aligned with directional changes; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Swift weight transfers with precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02697_slice17", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02697_slice17.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with sharp transitions; #head_movement: Quick tilts and rotations aligned with directional changes; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Swift weight transfers with precise footwork; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "window_end": 300, "window_start": 240}}
|
| 49 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4979, "index": 159, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02728_slice8.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: fluid arm waves combined with pops; #legs_movement: bent knees with staggered weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02728_slice8", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02728_slice8.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: fluid arm waves combined with pops; #legs_movement: bent knees with staggered weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 60, "window_start": 0}}
|
| 50 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4987, "index": 167, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__02820_slice2.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with strong directional changes; #head_movement: Quick tilts and turns in sync with torso; #arms_movement: Sharp, precise extensions and rapid repositioning; #legs_movement: Swift weight transfers with agile steps and jumps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__02820_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__02820_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic shifts with strong directional changes; #head_movement: Quick tilts and turns in sync with torso; #arms_movement: Sharp, precise extensions and rapid repositioning; #legs_movement: Swift weight transfers with agile steps and jumps; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium distance.", "window_end": 300, "window_start": 240}}
|
| 51 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4993, "index": 173, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__03030_slice3.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick tilts and pops in sync with beats; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Low stances with quick footwork and knee lifts; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__03030_slice3", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__03030_slice3.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick tilts and pops in sync with beats; #arms_movement: Rapid extensions and contractions, often asymmetrical; #legs_movement: Low stances with quick footwork and knee lifts; #camera_angle: Frontal view; #camera_distance: Full-body visible, medium shot.", "window_end": 300, "window_start": 240}}
|
| 52 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 4995, "index": 175, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__03081_slice5.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent-knee steps with weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__03081_slice5", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__03081_slice5.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent-knee steps with weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 180, "window_start": 120}}
|
| 53 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5003, "index": 183, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__03341_slice21.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: minimal, aligned with torso; #arms_movement: sharp and coordinated with upper body hits; #legs_movement: bent knees with low, bouncy stances; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__03341_slice21", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__03341_slice21.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: minimal, aligned with torso; #arms_movement: sharp and coordinated with upper body hits; #legs_movement: bent knees with low, bouncy stances; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 60, "window_start": 0}}
|
| 54 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5004, "index": 184, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__03341_slice21.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: minimal, aligned with torso; #arms_movement: sharp and coordinated with upper body hits; #legs_movement: bent knees with low, bouncy stances; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__03341_slice21", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__03341_slice21.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: minimal, aligned with torso; #arms_movement: sharp and coordinated with upper body hits; #legs_movement: bent knees with low, bouncy stances; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 180, "window_start": 120}}
|
| 55 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5009, "index": 189, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__03836_slice0.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: extended and punctuated with controlled swings; #legs_movement: bent knees with fluid steps and weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__03836_slice0", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__03836_slice0.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: extended and punctuated with controlled swings; #legs_movement: bent knees with fluid steps and weight shifts; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 60, "window_start": 0}}
|
| 56 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5030, "index": 210, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__05608_slice29.json", "segment_action": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp isolations; #head_movement: Quick directional changes aligned with torso; #arms_movement: Extended, precise gestures with rapid repositioning; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__05608_slice29", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__05608_slice29.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a National dance with modern style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic transitions with sharp isolations; #head_movement: Quick directional changes aligned with torso; #arms_movement: Extended, precise gestures with rapid repositioning; #legs_movement: Agile steps and controlled lifts; #camera_angle: Frontal view; #camera_distance: Full-body visibility.", "window_end": 60, "window_start": 0}}
|
| 57 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5036, "index": 216, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__05842_slice2.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick directional changes and tilts; #arms_movement: Rapid, precise extensions and contractions; #legs_movement: Agile steps and quick weight shifts; #camera_angle: Frontal view; #camera_distance: Full body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__05842_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__05842_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick directional changes and tilts; #arms_movement: Rapid, precise extensions and contractions; #legs_movement: Agile steps and quick weight shifts; #camera_angle: Frontal view; #camera_distance: Full body visible.", "window_end": 60, "window_start": 0}}
|
| 58 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5037, "index": 217, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__05842_slice2.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick directional changes and tilts; #arms_movement: Rapid, precise extensions and contractions; #legs_movement: Agile steps and quick weight shifts; #camera_angle: Frontal view; #camera_distance: Full body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__05842_slice2", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__05842_slice2.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: Fast and complicated; #emotional_expression: Intense and focused; #whole_body_movement: Dynamic isolations with sharp transitions; #head_movement: Quick directional changes and tilts; #arms_movement: Rapid, precise extensions and contractions; #legs_movement: Agile steps and quick weight shifts; #camera_angle: Frontal view; #camera_distance: Full body visible.", "window_end": 180, "window_start": 120}}
|
| 59 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5043, "index": 223, "meta": {"eval_num_windows": 3, "eval_window_index": 1, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__05873_slice25.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and gestural with fluid transitions; #legs_movement: bent knees with bouncy, weighted steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__05873_slice25", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__05873_slice25.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and gestural with fluid transitions; #legs_movement: bent knees with bouncy, weighted steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 180, "window_start": 120}}
|
| 60 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5050, "index": 230, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__05949_slice4.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid isolations and rhythmic undulations; #emotional_expression: confident and grounded intensity; #whole_body_movement: dynamic vertical shifts with strong downbeats; #head_movement: sharp tilts and bobs in sync with footwork; #arms_movement: alternating extensions and flexions accenting grooves; #legs_movement: bent-knee steps with controlled lateral weight transfers; #camera_angle: front-facing, eye-level; #camera_distance: full-body visible, medium shot.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__05949_slice4", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__05949_slice4.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: fluid isolations and rhythmic undulations; #emotional_expression: confident and grounded intensity; #whole_body_movement: dynamic vertical shifts with strong downbeats; #head_movement: sharp tilts and bobs in sync with footwork; #arms_movement: alternating extensions and flexions accenting grooves; #legs_movement: bent-knee steps with controlled lateral weight transfers; #camera_angle: front-facing, eye-level; #camera_distance: full-body visible, medium shot.", "window_end": 300, "window_start": 240}}
|
| 61 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5065, "index": 245, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__06211_slice0.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: punchy and angular with controlled extensions; #legs_movement: bent knees with weighted steps and quick shifts; #camera_angle: front-facing, capturing full body; #camera_distance: medium shot, focused on torso and limbs.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__06211_slice0", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__06211_slice0.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: punchy and angular with controlled extensions; #legs_movement: bent knees with weighted steps and quick shifts; #camera_angle: front-facing, capturing full body; #camera_distance: medium shot, focused on torso and limbs.", "window_end": 300, "window_start": 240}}
|
| 62 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5075, "index": 255, "meta": {"eval_num_windows": 3, "eval_window_index": 0, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__06631_slice1.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: fluid waves combined with punctuated poses; #legs_movement: bent knees with controlled steps and weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__06631_slice1", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__06631_slice1.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and intense; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp tilts and bobs in sync with beats; #arms_movement: fluid waves combined with punctuated poses; #legs_movement: bent knees with controlled steps and weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 60, "window_start": 0}}
|
| 63 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5083, "index": 263, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__06651_slice3.json", "segment_action": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent knees with bouncy, staggered steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__06651_slice3", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__06651_slice3.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a #genre Street dance with #detail_genre Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: extended and punctuated gestures; #legs_movement: bent knees with bouncy, staggered steps; #camera_angle: front-facing; #camera_distance: full-body visible.", "window_end": 300, "window_start": 240}}
|
| 64 |
+
{"caption": "", "dataset": "opendance_original_audio_test", "global_index": 5086, "index": 266, "meta": {"eval_num_windows": 3, "eval_window_index": 2, "fps": 30.0, "label_path": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/labels/test__06657_slice3.json", "segment_action": "The dancer is performing a Street dance with Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with subtle weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "segment_frame_end": 300, "segment_frame_start": 0, "segment_index": 0, "segment_style": "dance", "sequence_name": "test__06657_slice3", "source_file": "${OMG_STORAGE_ROOT}/filtered_datasets/original/opendance/g1/test__06657_slice3.npz", "split": "test", "text_path": null, "video_summary": "The dancer is performing a Street dance with Hiphop style. #movement_style: grounded and rhythmic; #emotional_expression: confident and controlled; #whole_body_movement: dynamic isolations with strong downbeats; #head_movement: sharp directional changes; #arms_movement: fluid yet punctuated gestures; #legs_movement: bent-knee stances with subtle weight shifts; #camera_angle: frontal view; #camera_distance: full-body visible.", "window_end": 300, "window_start": 240}}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_007/samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shards.json
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/audio_test_512.jsonl",
|
| 3 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards",
|
| 4 |
+
"num_samples": 512,
|
| 5 |
+
"num_shards": 8,
|
| 6 |
+
"shards": [
|
| 7 |
+
{
|
| 8 |
+
"shard_index": 0,
|
| 9 |
+
"start": 0,
|
| 10 |
+
"end": 64,
|
| 11 |
+
"num_samples": 64,
|
| 12 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_000/samples.jsonl",
|
| 13 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_000"
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"shard_index": 1,
|
| 17 |
+
"start": 64,
|
| 18 |
+
"end": 128,
|
| 19 |
+
"num_samples": 64,
|
| 20 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_001/samples.jsonl",
|
| 21 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_001"
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"shard_index": 2,
|
| 25 |
+
"start": 128,
|
| 26 |
+
"end": 192,
|
| 27 |
+
"num_samples": 64,
|
| 28 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_002/samples.jsonl",
|
| 29 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_002"
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"shard_index": 3,
|
| 33 |
+
"start": 192,
|
| 34 |
+
"end": 256,
|
| 35 |
+
"num_samples": 64,
|
| 36 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_003/samples.jsonl",
|
| 37 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_003"
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"shard_index": 4,
|
| 41 |
+
"start": 256,
|
| 42 |
+
"end": 320,
|
| 43 |
+
"num_samples": 64,
|
| 44 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_004/samples.jsonl",
|
| 45 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_004"
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"shard_index": 5,
|
| 49 |
+
"start": 320,
|
| 50 |
+
"end": 384,
|
| 51 |
+
"num_samples": 64,
|
| 52 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_005/samples.jsonl",
|
| 53 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_005"
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"shard_index": 6,
|
| 57 |
+
"start": 384,
|
| 58 |
+
"end": 448,
|
| 59 |
+
"num_samples": 64,
|
| 60 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_006/samples.jsonl",
|
| 61 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_006"
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"shard_index": 7,
|
| 65 |
+
"start": 448,
|
| 66 |
+
"end": 512,
|
| 67 |
+
"num_samples": 64,
|
| 68 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_007/samples.jsonl",
|
| 69 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/shards/shard_007"
|
| 70 |
+
}
|
| 71 |
+
]
|
| 72 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/summary.md
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Generation Benchmark Summary
|
| 2 |
+
|
| 3 |
+
- benchmark: artifact_audio
|
| 4 |
+
- label: OmniMotion 300M mixed-modal s055000
|
| 5 |
+
- exp: g1_transformer_guided_300m
|
| 6 |
+
- split: test
|
| 7 |
+
|
| 8 |
+
| ckpt | FID(seq) | KID(seq) | R@1 | R@2 | R@3 | Diversity | Matching | MultiModal Dist | BeatAlign | BeatAlign Ref | BeatAlign Gap | FIDk | FIDg | Divk | Divg | PFC | PFC Ref | contact_sliding | foot_ground_error | body_jerk | PJ | AUJ | Tracker g-MPJPE | Tracker MPJPE | Tracker E_vel | Tracker E_acc |
|
| 9 |
+
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
|
| 10 |
+
| OmniMotion 300M mixed-modal s055000 | N/A | N/A | N/A | N/A | N/A | N/A | N/A | N/A | **0.4938 +/- 0.1461** | **0.5613 +/- 0.1361** | **-0.0675 +/- 0.1887** | **59.1969** | **0.7417** | **10.6529** | **4.8664** | **0.5747 +/- 0.6584** | **0.4756 +/- 0.7533** | **0.4061 +/- 0.2216** | **0.0410 +/- 0.0284** | **95.7334 +/- 40.6350** | N/A | N/A | **95.1012 +/- 39.9806** | **41.0693 +/- 15.8622** | **6.1034 +/- 2.3868** | **1.8976 +/- 0.8003** |
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed.json
ADDED
|
@@ -0,0 +1,565 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_clip": 10.0,
|
| 3 |
+
"artifact_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_executed_qpos.npz",
|
| 4 |
+
"benchmark_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/benchmark_tracker.json",
|
| 5 |
+
"control_substeps": 10,
|
| 6 |
+
"enabled": true,
|
| 7 |
+
"holomotion_onnx": "${OMG_MODELS_ROOT}/HoloMotion/HoloMotion_v1.2/holomotion_v1.2_motion_tracking_model/exported/model.onnx",
|
| 8 |
+
"metrics": {
|
| 9 |
+
"e_acc": {
|
| 10 |
+
"max": 5.082680843770504,
|
| 11 |
+
"mean": 1.897563667569102,
|
| 12 |
+
"min": 0.35615143133327365,
|
| 13 |
+
"std": 0.8002793205380141
|
| 14 |
+
},
|
| 15 |
+
"e_vel": {
|
| 16 |
+
"max": 17.04840175807476,
|
| 17 |
+
"mean": 6.103366651132092,
|
| 18 |
+
"min": 1.4394153840839863,
|
| 19 |
+
"std": 2.3867622499218415
|
| 20 |
+
},
|
| 21 |
+
"enabled": true,
|
| 22 |
+
"g_mpjpe": {
|
| 23 |
+
"max": 341.753214597702,
|
| 24 |
+
"mean": 95.1012089572032,
|
| 25 |
+
"min": 28.030861169099808,
|
| 26 |
+
"std": 39.98059037197871
|
| 27 |
+
},
|
| 28 |
+
"mpjpe": {
|
| 29 |
+
"max": 144.38872039318085,
|
| 30 |
+
"mean": 41.06933371076593,
|
| 31 |
+
"min": 18.016833811998367,
|
| 32 |
+
"std": 15.862212133962863
|
| 33 |
+
},
|
| 34 |
+
"num_samples": 512,
|
| 35 |
+
"root_index": 0,
|
| 36 |
+
"target_fps": 50.0
|
| 37 |
+
},
|
| 38 |
+
"metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_metrics.json",
|
| 39 |
+
"num_generated_samples": 512,
|
| 40 |
+
"num_samples": 512,
|
| 41 |
+
"providers": "TensorrtExecutionProvider,CUDAExecutionProvider",
|
| 42 |
+
"robot_xml": null,
|
| 43 |
+
"root_index": 0,
|
| 44 |
+
"sample_metrics_path": "${OMG_REPO_ROOT}/outputs/benchmark_results_mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/tracker_executed/tracker_sample_metrics.jsonl",
|
| 45 |
+
"selected_indices": [
|
| 46 |
+
0,
|
| 47 |
+
1,
|
| 48 |
+
2,
|
| 49 |
+
3,
|
| 50 |
+
4,
|
| 51 |
+
5,
|
| 52 |
+
6,
|
| 53 |
+
7,
|
| 54 |
+
8,
|
| 55 |
+
9,
|
| 56 |
+
10,
|
| 57 |
+
11,
|
| 58 |
+
12,
|
| 59 |
+
13,
|
| 60 |
+
14,
|
| 61 |
+
15,
|
| 62 |
+
16,
|
| 63 |
+
17,
|
| 64 |
+
18,
|
| 65 |
+
19,
|
| 66 |
+
20,
|
| 67 |
+
21,
|
| 68 |
+
22,
|
| 69 |
+
23,
|
| 70 |
+
24,
|
| 71 |
+
25,
|
| 72 |
+
26,
|
| 73 |
+
27,
|
| 74 |
+
28,
|
| 75 |
+
29,
|
| 76 |
+
30,
|
| 77 |
+
31,
|
| 78 |
+
32,
|
| 79 |
+
33,
|
| 80 |
+
34,
|
| 81 |
+
35,
|
| 82 |
+
36,
|
| 83 |
+
37,
|
| 84 |
+
38,
|
| 85 |
+
39,
|
| 86 |
+
40,
|
| 87 |
+
41,
|
| 88 |
+
42,
|
| 89 |
+
43,
|
| 90 |
+
44,
|
| 91 |
+
45,
|
| 92 |
+
46,
|
| 93 |
+
47,
|
| 94 |
+
48,
|
| 95 |
+
49,
|
| 96 |
+
50,
|
| 97 |
+
51,
|
| 98 |
+
52,
|
| 99 |
+
53,
|
| 100 |
+
54,
|
| 101 |
+
55,
|
| 102 |
+
56,
|
| 103 |
+
57,
|
| 104 |
+
58,
|
| 105 |
+
59,
|
| 106 |
+
60,
|
| 107 |
+
61,
|
| 108 |
+
62,
|
| 109 |
+
63,
|
| 110 |
+
64,
|
| 111 |
+
65,
|
| 112 |
+
66,
|
| 113 |
+
67,
|
| 114 |
+
68,
|
| 115 |
+
69,
|
| 116 |
+
70,
|
| 117 |
+
71,
|
| 118 |
+
72,
|
| 119 |
+
73,
|
| 120 |
+
74,
|
| 121 |
+
75,
|
| 122 |
+
76,
|
| 123 |
+
77,
|
| 124 |
+
78,
|
| 125 |
+
79,
|
| 126 |
+
80,
|
| 127 |
+
81,
|
| 128 |
+
82,
|
| 129 |
+
83,
|
| 130 |
+
84,
|
| 131 |
+
85,
|
| 132 |
+
86,
|
| 133 |
+
87,
|
| 134 |
+
88,
|
| 135 |
+
89,
|
| 136 |
+
90,
|
| 137 |
+
91,
|
| 138 |
+
92,
|
| 139 |
+
93,
|
| 140 |
+
94,
|
| 141 |
+
95,
|
| 142 |
+
96,
|
| 143 |
+
97,
|
| 144 |
+
98,
|
| 145 |
+
99,
|
| 146 |
+
100,
|
| 147 |
+
101,
|
| 148 |
+
102,
|
| 149 |
+
103,
|
| 150 |
+
104,
|
| 151 |
+
105,
|
| 152 |
+
106,
|
| 153 |
+
107,
|
| 154 |
+
108,
|
| 155 |
+
109,
|
| 156 |
+
110,
|
| 157 |
+
111,
|
| 158 |
+
112,
|
| 159 |
+
113,
|
| 160 |
+
114,
|
| 161 |
+
115,
|
| 162 |
+
116,
|
| 163 |
+
117,
|
| 164 |
+
118,
|
| 165 |
+
119,
|
| 166 |
+
120,
|
| 167 |
+
121,
|
| 168 |
+
122,
|
| 169 |
+
123,
|
| 170 |
+
124,
|
| 171 |
+
125,
|
| 172 |
+
126,
|
| 173 |
+
127,
|
| 174 |
+
128,
|
| 175 |
+
129,
|
| 176 |
+
130,
|
| 177 |
+
131,
|
| 178 |
+
132,
|
| 179 |
+
133,
|
| 180 |
+
134,
|
| 181 |
+
135,
|
| 182 |
+
136,
|
| 183 |
+
137,
|
| 184 |
+
138,
|
| 185 |
+
139,
|
| 186 |
+
140,
|
| 187 |
+
141,
|
| 188 |
+
142,
|
| 189 |
+
143,
|
| 190 |
+
144,
|
| 191 |
+
145,
|
| 192 |
+
146,
|
| 193 |
+
147,
|
| 194 |
+
148,
|
| 195 |
+
149,
|
| 196 |
+
150,
|
| 197 |
+
151,
|
| 198 |
+
152,
|
| 199 |
+
153,
|
| 200 |
+
154,
|
| 201 |
+
155,
|
| 202 |
+
156,
|
| 203 |
+
157,
|
| 204 |
+
158,
|
| 205 |
+
159,
|
| 206 |
+
160,
|
| 207 |
+
161,
|
| 208 |
+
162,
|
| 209 |
+
163,
|
| 210 |
+
164,
|
| 211 |
+
165,
|
| 212 |
+
166,
|
| 213 |
+
167,
|
| 214 |
+
168,
|
| 215 |
+
169,
|
| 216 |
+
170,
|
| 217 |
+
171,
|
| 218 |
+
172,
|
| 219 |
+
173,
|
| 220 |
+
174,
|
| 221 |
+
175,
|
| 222 |
+
176,
|
| 223 |
+
177,
|
| 224 |
+
178,
|
| 225 |
+
179,
|
| 226 |
+
180,
|
| 227 |
+
181,
|
| 228 |
+
182,
|
| 229 |
+
183,
|
| 230 |
+
184,
|
| 231 |
+
185,
|
| 232 |
+
186,
|
| 233 |
+
187,
|
| 234 |
+
188,
|
| 235 |
+
189,
|
| 236 |
+
190,
|
| 237 |
+
191,
|
| 238 |
+
192,
|
| 239 |
+
193,
|
| 240 |
+
194,
|
| 241 |
+
195,
|
| 242 |
+
196,
|
| 243 |
+
197,
|
| 244 |
+
198,
|
| 245 |
+
199,
|
| 246 |
+
200,
|
| 247 |
+
201,
|
| 248 |
+
202,
|
| 249 |
+
203,
|
| 250 |
+
204,
|
| 251 |
+
205,
|
| 252 |
+
206,
|
| 253 |
+
207,
|
| 254 |
+
208,
|
| 255 |
+
209,
|
| 256 |
+
210,
|
| 257 |
+
211,
|
| 258 |
+
212,
|
| 259 |
+
213,
|
| 260 |
+
214,
|
| 261 |
+
215,
|
| 262 |
+
216,
|
| 263 |
+
217,
|
| 264 |
+
218,
|
| 265 |
+
219,
|
| 266 |
+
220,
|
| 267 |
+
221,
|
| 268 |
+
222,
|
| 269 |
+
223,
|
| 270 |
+
224,
|
| 271 |
+
225,
|
| 272 |
+
226,
|
| 273 |
+
227,
|
| 274 |
+
228,
|
| 275 |
+
229,
|
| 276 |
+
230,
|
| 277 |
+
231,
|
| 278 |
+
232,
|
| 279 |
+
233,
|
| 280 |
+
234,
|
| 281 |
+
235,
|
| 282 |
+
236,
|
| 283 |
+
237,
|
| 284 |
+
238,
|
| 285 |
+
239,
|
| 286 |
+
240,
|
| 287 |
+
241,
|
| 288 |
+
242,
|
| 289 |
+
243,
|
| 290 |
+
244,
|
| 291 |
+
245,
|
| 292 |
+
246,
|
| 293 |
+
247,
|
| 294 |
+
248,
|
| 295 |
+
249,
|
| 296 |
+
250,
|
| 297 |
+
251,
|
| 298 |
+
252,
|
| 299 |
+
253,
|
| 300 |
+
254,
|
| 301 |
+
255,
|
| 302 |
+
256,
|
| 303 |
+
257,
|
| 304 |
+
258,
|
| 305 |
+
259,
|
| 306 |
+
260,
|
| 307 |
+
261,
|
| 308 |
+
262,
|
| 309 |
+
263,
|
| 310 |
+
264,
|
| 311 |
+
265,
|
| 312 |
+
266,
|
| 313 |
+
267,
|
| 314 |
+
268,
|
| 315 |
+
269,
|
| 316 |
+
270,
|
| 317 |
+
271,
|
| 318 |
+
272,
|
| 319 |
+
273,
|
| 320 |
+
274,
|
| 321 |
+
275,
|
| 322 |
+
276,
|
| 323 |
+
277,
|
| 324 |
+
278,
|
| 325 |
+
279,
|
| 326 |
+
280,
|
| 327 |
+
281,
|
| 328 |
+
282,
|
| 329 |
+
283,
|
| 330 |
+
284,
|
| 331 |
+
285,
|
| 332 |
+
286,
|
| 333 |
+
287,
|
| 334 |
+
288,
|
| 335 |
+
289,
|
| 336 |
+
290,
|
| 337 |
+
291,
|
| 338 |
+
292,
|
| 339 |
+
293,
|
| 340 |
+
294,
|
| 341 |
+
295,
|
| 342 |
+
296,
|
| 343 |
+
297,
|
| 344 |
+
298,
|
| 345 |
+
299,
|
| 346 |
+
300,
|
| 347 |
+
301,
|
| 348 |
+
302,
|
| 349 |
+
303,
|
| 350 |
+
304,
|
| 351 |
+
305,
|
| 352 |
+
306,
|
| 353 |
+
307,
|
| 354 |
+
308,
|
| 355 |
+
309,
|
| 356 |
+
310,
|
| 357 |
+
311,
|
| 358 |
+
312,
|
| 359 |
+
313,
|
| 360 |
+
314,
|
| 361 |
+
315,
|
| 362 |
+
316,
|
| 363 |
+
317,
|
| 364 |
+
318,
|
| 365 |
+
319,
|
| 366 |
+
320,
|
| 367 |
+
321,
|
| 368 |
+
322,
|
| 369 |
+
323,
|
| 370 |
+
324,
|
| 371 |
+
325,
|
| 372 |
+
326,
|
| 373 |
+
327,
|
| 374 |
+
328,
|
| 375 |
+
329,
|
| 376 |
+
330,
|
| 377 |
+
331,
|
| 378 |
+
332,
|
| 379 |
+
333,
|
| 380 |
+
334,
|
| 381 |
+
335,
|
| 382 |
+
336,
|
| 383 |
+
337,
|
| 384 |
+
338,
|
| 385 |
+
339,
|
| 386 |
+
340,
|
| 387 |
+
341,
|
| 388 |
+
342,
|
| 389 |
+
343,
|
| 390 |
+
344,
|
| 391 |
+
345,
|
| 392 |
+
346,
|
| 393 |
+
347,
|
| 394 |
+
348,
|
| 395 |
+
349,
|
| 396 |
+
350,
|
| 397 |
+
351,
|
| 398 |
+
352,
|
| 399 |
+
353,
|
| 400 |
+
354,
|
| 401 |
+
355,
|
| 402 |
+
356,
|
| 403 |
+
357,
|
| 404 |
+
358,
|
| 405 |
+
359,
|
| 406 |
+
360,
|
| 407 |
+
361,
|
| 408 |
+
362,
|
| 409 |
+
363,
|
| 410 |
+
364,
|
| 411 |
+
365,
|
| 412 |
+
366,
|
| 413 |
+
367,
|
| 414 |
+
368,
|
| 415 |
+
369,
|
| 416 |
+
370,
|
| 417 |
+
371,
|
| 418 |
+
372,
|
| 419 |
+
373,
|
| 420 |
+
374,
|
| 421 |
+
375,
|
| 422 |
+
376,
|
| 423 |
+
377,
|
| 424 |
+
378,
|
| 425 |
+
379,
|
| 426 |
+
380,
|
| 427 |
+
381,
|
| 428 |
+
382,
|
| 429 |
+
383,
|
| 430 |
+
384,
|
| 431 |
+
385,
|
| 432 |
+
386,
|
| 433 |
+
387,
|
| 434 |
+
388,
|
| 435 |
+
389,
|
| 436 |
+
390,
|
| 437 |
+
391,
|
| 438 |
+
392,
|
| 439 |
+
393,
|
| 440 |
+
394,
|
| 441 |
+
395,
|
| 442 |
+
396,
|
| 443 |
+
397,
|
| 444 |
+
398,
|
| 445 |
+
399,
|
| 446 |
+
400,
|
| 447 |
+
401,
|
| 448 |
+
402,
|
| 449 |
+
403,
|
| 450 |
+
404,
|
| 451 |
+
405,
|
| 452 |
+
406,
|
| 453 |
+
407,
|
| 454 |
+
408,
|
| 455 |
+
409,
|
| 456 |
+
410,
|
| 457 |
+
411,
|
| 458 |
+
412,
|
| 459 |
+
413,
|
| 460 |
+
414,
|
| 461 |
+
415,
|
| 462 |
+
416,
|
| 463 |
+
417,
|
| 464 |
+
418,
|
| 465 |
+
419,
|
| 466 |
+
420,
|
| 467 |
+
421,
|
| 468 |
+
422,
|
| 469 |
+
423,
|
| 470 |
+
424,
|
| 471 |
+
425,
|
| 472 |
+
426,
|
| 473 |
+
427,
|
| 474 |
+
428,
|
| 475 |
+
429,
|
| 476 |
+
430,
|
| 477 |
+
431,
|
| 478 |
+
432,
|
| 479 |
+
433,
|
| 480 |
+
434,
|
| 481 |
+
435,
|
| 482 |
+
436,
|
| 483 |
+
437,
|
| 484 |
+
438,
|
| 485 |
+
439,
|
| 486 |
+
440,
|
| 487 |
+
441,
|
| 488 |
+
442,
|
| 489 |
+
443,
|
| 490 |
+
444,
|
| 491 |
+
445,
|
| 492 |
+
446,
|
| 493 |
+
447,
|
| 494 |
+
448,
|
| 495 |
+
449,
|
| 496 |
+
450,
|
| 497 |
+
451,
|
| 498 |
+
452,
|
| 499 |
+
453,
|
| 500 |
+
454,
|
| 501 |
+
455,
|
| 502 |
+
456,
|
| 503 |
+
457,
|
| 504 |
+
458,
|
| 505 |
+
459,
|
| 506 |
+
460,
|
| 507 |
+
461,
|
| 508 |
+
462,
|
| 509 |
+
463,
|
| 510 |
+
464,
|
| 511 |
+
465,
|
| 512 |
+
466,
|
| 513 |
+
467,
|
| 514 |
+
468,
|
| 515 |
+
469,
|
| 516 |
+
470,
|
| 517 |
+
471,
|
| 518 |
+
472,
|
| 519 |
+
473,
|
| 520 |
+
474,
|
| 521 |
+
475,
|
| 522 |
+
476,
|
| 523 |
+
477,
|
| 524 |
+
478,
|
| 525 |
+
479,
|
| 526 |
+
480,
|
| 527 |
+
481,
|
| 528 |
+
482,
|
| 529 |
+
483,
|
| 530 |
+
484,
|
| 531 |
+
485,
|
| 532 |
+
486,
|
| 533 |
+
487,
|
| 534 |
+
488,
|
| 535 |
+
489,
|
| 536 |
+
490,
|
| 537 |
+
491,
|
| 538 |
+
492,
|
| 539 |
+
493,
|
| 540 |
+
494,
|
| 541 |
+
495,
|
| 542 |
+
496,
|
| 543 |
+
497,
|
| 544 |
+
498,
|
| 545 |
+
499,
|
| 546 |
+
500,
|
| 547 |
+
501,
|
| 548 |
+
502,
|
| 549 |
+
503,
|
| 550 |
+
504,
|
| 551 |
+
505,
|
| 552 |
+
506,
|
| 553 |
+
507,
|
| 554 |
+
508,
|
| 555 |
+
509,
|
| 556 |
+
510,
|
| 557 |
+
511
|
| 558 |
+
],
|
| 559 |
+
"steps": null,
|
| 560 |
+
"target_fps": 50.0,
|
| 561 |
+
"valid_shape": [
|
| 562 |
+
512,
|
| 563 |
+
99
|
| 564 |
+
]
|
| 565 |
+
}
|
benchmark/results/mixed_modalities_all_v1/audio/omnimotion_300m_mixed_modal_all_s055000/transition_metrics.json
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"chunk_length": 60,
|
| 3 |
+
"enabled": false,
|
| 4 |
+
"reason": "disabled"
|
| 5 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/.audio_120f_export/audio/manifest.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/.audio_120f_export/export_summary.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/audio_test_512.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/audio_test_512_120f.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/gmr_smplx/smplx/manifest.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/gmr_smplx_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/keypoints_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/materialization_summary.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"num_frames": 60,
|
| 3 |
+
"num_samples": 512,
|
| 4 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs",
|
| 5 |
+
"outputs": {
|
| 6 |
+
"keypoints_samples": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs/keypoints_samples.jsonl",
|
| 7 |
+
"phc_amass": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs/phc_input_samples.jsonl"
|
| 8 |
+
},
|
| 9 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_test_512.jsonl",
|
| 10 |
+
"stages": [
|
| 11 |
+
"keypoints",
|
| 12 |
+
"phc_amass"
|
| 13 |
+
]
|
| 14 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/nmr_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/omniretarget_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs/phc_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_phc_zup_fixed/phc_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/keypoints_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/materialization_summary.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"num_frames": 60,
|
| 3 |
+
"num_samples": 512,
|
| 4 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup",
|
| 5 |
+
"outputs": {
|
| 6 |
+
"keypoints_samples": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup/keypoints_samples.jsonl",
|
| 7 |
+
"nmr_smplx": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup/nmr_input_samples.jsonl"
|
| 8 |
+
},
|
| 9 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_test_512.jsonl",
|
| 10 |
+
"stages": [
|
| 11 |
+
"keypoints",
|
| 12 |
+
"nmr_smplx"
|
| 13 |
+
]
|
| 14 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup/nmr_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/gmr_smplx/smplx/manifest.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/gmr_smplx_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/keypoints_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/materialization_summary.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"num_frames": 120,
|
| 3 |
+
"num_samples": 512,
|
| 4 |
+
"output_dir": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup_120f",
|
| 5 |
+
"outputs": {
|
| 6 |
+
"keypoints_samples": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/keypoints_samples.jsonl",
|
| 7 |
+
"omniretarget_motion_dir": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/omniretarget_smplx"
|
| 8 |
+
},
|
| 9 |
+
"samples_path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_test_512_120f.jsonl",
|
| 10 |
+
"stages": [
|
| 11 |
+
"omniretarget"
|
| 12 |
+
]
|
| 13 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/nmr_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_inputs_zup_120f/omniretarget_input_samples.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_test_512.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/humanref_test_512_120f.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/manifest_summary.json
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark_sample_set": "mixed_modalities_all_v1",
|
| 3 |
+
"conditions": {
|
| 4 |
+
"audio": {
|
| 5 |
+
"num_samples": 512,
|
| 6 |
+
"path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/audio_test_512.jsonl",
|
| 7 |
+
"selected_counts": {
|
| 8 |
+
"aioz_gdance_original_audio_test": 103,
|
| 9 |
+
"aistpp_original_audio_test": 103,
|
| 10 |
+
"compas3d_original_audio_test": 102,
|
| 11 |
+
"finedance_original_audio_test": 102,
|
| 12 |
+
"opendance_original_audio_test": 102
|
| 13 |
+
}
|
| 14 |
+
},
|
| 15 |
+
"humanref": {
|
| 16 |
+
"num_samples": 512,
|
| 17 |
+
"path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_test_512.jsonl",
|
| 18 |
+
"selected_counts": {
|
| 19 |
+
"aistpp_original_audio_test": 47,
|
| 20 |
+
"amass_original_filtered_test": 47,
|
| 21 |
+
"beat2_original_humanref_test": 47,
|
| 22 |
+
"finedance_original_audio_test": 47,
|
| 23 |
+
"fitness_original_filtered_test": 47,
|
| 24 |
+
"humanml_original_filtered_test": 47,
|
| 25 |
+
"idea400_original_filtered_test": 46,
|
| 26 |
+
"motiongv_original_filtered_test": 46,
|
| 27 |
+
"motionllama_original_filtered_test": 46,
|
| 28 |
+
"permo_original_filtered_test": 46,
|
| 29 |
+
"snapmogen_original_filtered_test": 46
|
| 30 |
+
}
|
| 31 |
+
},
|
| 32 |
+
"text": {
|
| 33 |
+
"num_samples": 1024,
|
| 34 |
+
"path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/text_test_1024.jsonl",
|
| 35 |
+
"selected_counts": {
|
| 36 |
+
"100style_original_filtered_test": 86,
|
| 37 |
+
"amass_original_filtered_test": 86,
|
| 38 |
+
"bones_seed_original_filtered_test": 86,
|
| 39 |
+
"fitness_original_filtered_test": 86,
|
| 40 |
+
"humanml_original_filtered_test": 85,
|
| 41 |
+
"idea400_original_filtered_test": 85,
|
| 42 |
+
"lafan1_original_filtered_test": 85,
|
| 43 |
+
"motiongv_original_filtered_test": 85,
|
| 44 |
+
"motionllama_original_filtered_test": 85,
|
| 45 |
+
"omomo_original_filtered_test": 85,
|
| 46 |
+
"permo_original_filtered_test": 85,
|
| 47 |
+
"snapmogen_original_filtered_test": 85
|
| 48 |
+
}
|
| 49 |
+
}
|
| 50 |
+
},
|
| 51 |
+
"data": "filtered_original_mixed_modalities_all",
|
| 52 |
+
"exp": "g1_transformer_guided_100m",
|
| 53 |
+
"num_frames": 60,
|
| 54 |
+
"seed": 0,
|
| 55 |
+
"split": "test"
|
| 56 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/text_test_1024.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/text_test_512_120f.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/samples/mixed_modalities_all_v1/text_transition_manifest_summary.json
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark_sample_set": "mixed_modalities_all_v1",
|
| 3 |
+
"conditions": {
|
| 4 |
+
"text": {
|
| 5 |
+
"candidate_counts": {
|
| 6 |
+
"100style_original_filtered_test": 3041,
|
| 7 |
+
"amass_original_filtered_test": 3586,
|
| 8 |
+
"bones_seed_original_filtered_test": 19375,
|
| 9 |
+
"fitness_original_filtered_test": 93,
|
| 10 |
+
"humanml_original_filtered_test": 6613,
|
| 11 |
+
"idea400_original_filtered_test": 2984,
|
| 12 |
+
"lafan1_original_filtered_test": 392,
|
| 13 |
+
"motiongv_original_filtered_test": 63370,
|
| 14 |
+
"motionllama_original_filtered_test": 1279,
|
| 15 |
+
"omomo_original_filtered_test": 100,
|
| 16 |
+
"permo_original_filtered_test": 1040,
|
| 17 |
+
"snapmogen_original_filtered_test": 11668
|
| 18 |
+
},
|
| 19 |
+
"num_samples": 512,
|
| 20 |
+
"path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/text_test_512_120f.jsonl",
|
| 21 |
+
"selected_counts": {
|
| 22 |
+
"100style_original_filtered_test": 43,
|
| 23 |
+
"amass_original_filtered_test": 43,
|
| 24 |
+
"bones_seed_original_filtered_test": 43,
|
| 25 |
+
"fitness_original_filtered_test": 43,
|
| 26 |
+
"humanml_original_filtered_test": 43,
|
| 27 |
+
"idea400_original_filtered_test": 43,
|
| 28 |
+
"lafan1_original_filtered_test": 43,
|
| 29 |
+
"motiongv_original_filtered_test": 43,
|
| 30 |
+
"motionllama_original_filtered_test": 42,
|
| 31 |
+
"omomo_original_filtered_test": 42,
|
| 32 |
+
"permo_original_filtered_test": 42,
|
| 33 |
+
"snapmogen_original_filtered_test": 42
|
| 34 |
+
}
|
| 35 |
+
}
|
| 36 |
+
},
|
| 37 |
+
"data": "filtered_original_mixed_modalities_all",
|
| 38 |
+
"exp": "g1_transformer_guided_300m",
|
| 39 |
+
"num_frames": 120,
|
| 40 |
+
"seed": 0,
|
| 41 |
+
"split": "test"
|
| 42 |
+
}
|
benchmark/samples/mixed_modalities_all_v1/transition_manifest_summary.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark_sample_set": "mixed_modalities_all_v1",
|
| 3 |
+
"conditions": {
|
| 4 |
+
"humanref": {
|
| 5 |
+
"candidate_counts": {
|
| 6 |
+
"aistpp_original_audio_test": 1986,
|
| 7 |
+
"amass_original_filtered_test": 3586,
|
| 8 |
+
"beat2_original_humanref_test": 126,
|
| 9 |
+
"finedance_original_audio_test": 1764,
|
| 10 |
+
"fitness_original_filtered_test": 93,
|
| 11 |
+
"humanml_original_filtered_test": 6613,
|
| 12 |
+
"idea400_original_filtered_test": 2984,
|
| 13 |
+
"motiongv_original_filtered_test": 63370,
|
| 14 |
+
"motionllama_original_filtered_test": 1279,
|
| 15 |
+
"permo_original_filtered_test": 1040,
|
| 16 |
+
"snapmogen_original_filtered_test": 11668
|
| 17 |
+
},
|
| 18 |
+
"num_samples": 512,
|
| 19 |
+
"path": "${OMG_REPO_ROOT}/outputs/benchmark_samples/mixed_modalities_all_v1/humanref_test_512_120f.jsonl",
|
| 20 |
+
"selected_counts": {
|
| 21 |
+
"aistpp_original_audio_test": 47,
|
| 22 |
+
"amass_original_filtered_test": 47,
|
| 23 |
+
"beat2_original_humanref_test": 47,
|
| 24 |
+
"finedance_original_audio_test": 47,
|
| 25 |
+
"fitness_original_filtered_test": 47,
|
| 26 |
+
"humanml_original_filtered_test": 47,
|
| 27 |
+
"idea400_original_filtered_test": 46,
|
| 28 |
+
"motiongv_original_filtered_test": 46,
|
| 29 |
+
"motionllama_original_filtered_test": 46,
|
| 30 |
+
"permo_original_filtered_test": 46,
|
| 31 |
+
"snapmogen_original_filtered_test": 46
|
| 32 |
+
}
|
| 33 |
+
}
|
| 34 |
+
},
|
| 35 |
+
"data": "filtered_original_mixed_modalities_all",
|
| 36 |
+
"exp": "g1_transformer_guided_100m",
|
| 37 |
+
"num_frames": 120,
|
| 38 |
+
"seed": 0,
|
| 39 |
+
"split": "test"
|
| 40 |
+
}
|