Datasets:
Tasks:
Visual Question Answering
Formats:
parquet
Languages:
English
Size:
100K - 1M
Tags:
video
multimodal
event-understanding
temporal-grounding
hierarchical-annotations
post-training
License:
Download metadata/sources.json from Lostgreen/EventLens-Train: direct link, hf CLI and curl.
- Browser
- Download file 651 Bytes
-
https://huggingface.co/datasets/Lostgreen/EventLens-Train/resolve/main/metadata/sources.json
- Command line
-
hf download hf://datasets/Lostgreen/EventLens-Train/metadata/sources.json
-
curl -L -o sources.json https://huggingface.co/datasets/Lostgreen/EventLens-Train/resolve/main/metadata/sources.json
651 Bytes
| { | |
| "et_instruct": { | |
| "url": "https://huggingface.co/datasets/PolyU-ChenLab/ET-Instruct-164K", | |
| "license": "CC-BY-NC-SA-4.0" | |
| }, | |
| "timelens": { | |
| "url": "https://huggingface.co/datasets/TencentARC/TimeLens-100K", | |
| "license": "TimeLens custom terms", | |
| "license_url": "https://github.com/TencentARC/TimeLens/blob/main/LICENSE" | |
| }, | |
| "timerft": { | |
| "url": "https://github.com/xiaomi-research/time-r1", | |
| "license": "Upstream TimeR1 and constituent dataset terms" | |
| }, | |
| "llava_video": { | |
| "url": "https://huggingface.co/datasets/lmms-lab/LLaVA-Video-178K", | |
| "license": "Upstream LLaVA-Video and constituent dataset terms" | |
| } | |
| } | |