Update eval_assets: GE30 list + yaw_v6cond PDM ckpt + README

README.md

@@ -1,26 +1,50 @@
-# Evaluation Assets Bundle
+# Evaluation Assets (GE30 + v6cond)
 
-This folder mirrors repository-relative paths for assets needed by
-`evaluation/eval_pool.py`.
+This folder is a **minimal Hugging Face upload payload** for updating
+[`UCBProject/eval_assets`](https://huggingface.co/datasets/UCBProject/eval_assets/tree/main).
 
-Included:
+It mirrors repository-relative paths, so after downloading you can sync it into
+the repo root and the files land in the correct locations.
+
+## What this update contains
+
+- `evaluation/configs/eval_objects_merged_success_ge30.csv`
+  - **87 objects** (criterion: round≥3 `robot_gt` cumulative `success_count ≥ 30`)
+- `evaluation/configs/README.md`
+  - Notes about GE30 being current; GE20/GE40 deprecated
+- `output/pdm/checkpoints_yaw_v6cond/best_model.pth`
+- `output/pdm/checkpoints_yaw_v6cond/pose_stats.pt`
+- `output/affordance_no_rot_executed/min20/checkpoints_v6/best_v6_model.pth`
+
+## What is intentionally NOT included
+
+This upload does **not** change heavy assets (mesh/USD/scene files), and does
+**not** include evaluation outputs:
 
 - `data_hub/meshes/SAM3DMesh/rotated_mesh/`
-- `data_hub/ProcessedData/obj_meshes/{oakink,ycb}/{obj_id}/scale.json` only (metric scale metadata; no mesh.ply / rotation.json)
+- `data_hub/ProcessedData/obj_meshes/*/scale.json`
 - `output/obj_usd/`
-- `output/affordance_no_rot_executed/min20/checkpoints_v6/best_v6_model.pth`
-- `output/pdm/checkpoints_yaw/best_model.pth`
-- `sim/assets_scene/`
-- `sim/object_rotation_overrides.json`
-- `evaluation/configs/eval_objects_merged_success_ge40.csv`
+- `sim/assets_scene/`, `sim/object_rotation_overrides.json`
+- `output/evaluation/**` (logs/videos/results)
+
+Those stay as-is in the existing dataset version.
+
+## How to use after download
+
+Download to a temp directory and sync into the repo root:
+
+```bash
+huggingface-cli download UCBProject/eval_assets \
+  --repo-type dataset \
+  --local-dir /tmp/a2g_eval_assets
+
+cd /path/to/Affordance2Grasp
+rsync -a /tmp/a2g_eval_assets/ ./
+```
 
-Not included:
+## Notes
 
-- cuRobo installation/source
-- Isaac Sim installation
-- pre-generated candidate HDF5s
-- merged robot GT files
-- evaluation outputs/videos/logs
+- The evaluation object list is now **GE30** (`eval_objects_merged_success_ge30.csv`).
+- PDM default checkpoint used by current eval scripts is **yaw-conditioned v6cond**
+  under `output/pdm/checkpoints_yaw_v6cond/`.
 
-After downloading the dataset, copy or sync its contents into the repository
-root so these relative paths exist.

evaluation/configs/README.md

@@ -0,0 +1,28 @@
+# Evaluation object lists
+
+## Current: `eval_objects_merged_success_ge30.csv`
+
+- **Criterion:** sum of `successful_grasps` in `output/grasp_collect_no_rot/robot_gt/round_R/` for **R ≥ 3**, per object, with **count ≥ 30**.
+- **Regenerate:**
+
+```bash
+python tools/build_eval_object_list.py \
+  --outdir output/grasp_collect_no_rot \
+  --min-round 3 \
+  --min-success 30 \
+  --output evaluation/configs/eval_objects_merged_success_ge30.csv
+```
+
+## Sim USD bundle
+
+Packed assets for external Isaac Sim eval:
+
+```bash
+python tools/package_eval_ge30_usd_bundle.py
+# → output/eval_objects_ge30_sim_bundle/
+```
+
+## Deprecated lists
+
+- **`eval_objects_merged_success_ge20.csv`** — round≥3, success≥20 (97 objects).
+- **`eval_objects_merged_success_ge40.csv`** — merged HDF5 total success≥40 (legacy rule).

evaluation/configs/eval_objects_merged_success_ge30.csv

@@ -0,0 +1,88 @@
+obj_id,enabled,success_count,notes
+A02015,1,259,round_ge3_success_ge30
+A02021,1,256,round_ge3_success_ge30
+A02028,1,230,round_ge3_success_ge30
+A02031,1,219,round_ge3_success_ge30
+A02014,1,211,round_ge3_success_ge30
+A01006,1,204,round_ge3_success_ge30
+A02030,1,195,round_ge3_success_ge30
+A15015,1,193,round_ge3_success_ge30
+O03002,1,187,round_ge3_success_ge30
+S15004,1,181,round_ge3_success_ge30
+A01026,1,167,round_ge3_success_ge30
+S16001,1,160,round_ge3_success_ge30
+O02001,1,123,round_ge3_success_ge30
+A01008,1,120,round_ge3_success_ge30
+S16002,1,120,round_ge3_success_ge30
+C14001,1,117,round_ge3_success_ge30
+A01023,1,116,round_ge3_success_ge30
+S16005,1,111,round_ge3_success_ge30
+O36001,1,110,round_ge3_success_ge30
+A01005,1,91,round_ge3_success_ge30
+A02018,1,91,round_ge3_success_ge30
+S10008,1,90,round_ge3_success_ge30
+ycb_dex_12,1,89,round_ge3_success_ge30
+Y03021,1,82,round_ge3_success_ge30
+A01027,1,80,round_ge3_success_ge30
+ycb_dex_04,1,79,round_ge3_success_ge30
+ycb_dex_07,1,72,round_ge3_success_ge30
+ycb_dex_02,1,71,round_ge3_success_ge30
+A01001,1,69,round_ge3_success_ge30
+C28001,1,69,round_ge3_success_ge30
+O01000,1,67,round_ge3_success_ge30
+A01010,1,64,round_ge3_success_ge30
+ycb_dex_08,1,63,round_ge3_success_ge30
+S10018,1,61,round_ge3_success_ge30
+A01002,1,60,round_ge3_success_ge30
+C50001,1,59,round_ge3_success_ge30
+O21001,1,59,round_ge3_success_ge30
+S20021,1,58,round_ge3_success_ge30
+A16026,1,57,round_ge3_success_ge30
+ycb_dex_03,1,57,round_ge3_success_ge30
+S10022,1,55,round_ge3_success_ge30
+S10011,1,54,round_ge3_success_ge30
+Y03006,1,52,round_ge3_success_ge30
+Y29040,1,51,round_ge3_success_ge30
+C90001,1,50,round_ge3_success_ge30
+S16003,1,50,round_ge3_success_ge30
+ycb_dex_14,1,50,round_ge3_success_ge30
+C51001,1,48,round_ge3_success_ge30
+O03001,1,48,round_ge3_success_ge30
+O36002,1,48,round_ge3_success_ge30
+S20022,1,48,round_ge3_success_ge30
+Y27035,1,47,round_ge3_success_ge30
+A16013,1,46,round_ge3_success_ge30
+S20001,1,46,round_ge3_success_ge30
+A01009,1,45,round_ge3_success_ge30
+A02011,1,45,round_ge3_success_ge30
+C15001,1,45,round_ge3_success_ge30
+O50001,1,45,round_ge3_success_ge30
+ycb_dex_05,1,44,round_ge3_success_ge30
+C22001,1,43,round_ge3_success_ge30
+S10017,1,43,round_ge3_success_ge30
+ycb_dex_17,1,43,round_ge3_success_ge30
+O03003,1,42,round_ge3_success_ge30
+A15027,1,41,round_ge3_success_ge30
+Y35037,1,41,round_ge3_success_ge30
+S10013,1,40,round_ge3_success_ge30
+S10010,1,39,round_ge3_success_ge30
+ycb_dex_06,1,39,round_ge3_success_ge30
+O93001,1,38,round_ge3_success_ge30
+ycb_dex_11,1,38,round_ge3_success_ge30
+A02029,1,37,round_ge3_success_ge30
+ycb_dex_01,1,37,round_ge3_success_ge30
+ycb_dex_18,1,37,round_ge3_success_ge30
+A02012,1,36,round_ge3_success_ge30
+S10005,1,36,round_ge3_success_ge30
+S10020,1,36,round_ge3_success_ge30
+O50002,1,35,round_ge3_success_ge30
+A16012,1,34,round_ge3_success_ge30
+C40001,1,33,round_ge3_success_ge30
+C41001,1,32,round_ge3_success_ge30
+S10014,1,32,round_ge3_success_ge30
+ycb_dex_15,1,32,round_ge3_success_ge30
+C03001,1,31,round_ge3_success_ge30
+C52001,1,31,round_ge3_success_ge30
+C42001,1,30,round_ge3_success_ge30
+C43001,1,30,round_ge3_success_ge30
+S10019,1,30,round_ge3_success_ge30

output/pdm/checkpoints_yaw_v6cond/best_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:803b6ecc6b8db2f00fbbce6f35a1c9fa1dd78f977a0a9d5758748740bd2c7535
+size 3673698

output/pdm/checkpoints_yaw_v6cond/pose_stats.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61c75e2d75632cd3c6c5b9bd082fbe0d3e4be2b3ac5e23d297ca216c02cd8703
+size 1450