Research map / Video / Action recognition
HA-ViD: A Human Assembly Video Dataset for Comprehensive Assembly Knowledge Understanding
Hao Zheng, Regina Lee, Yuqian Lu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fuse features from three camera views instead of single view (ablation) | BCN · HA-ViD action segmentation, primitive task | Acc: 44 -> 47.2 | Improved |
| fuse features from three camera views instead of single view (ablation) | MViTv2 · HA-ViD action recognition, atomic action | Top-1: 46.6 -> 50.8 | Improved |
| fuse features from three camera views instead of single view (ablation) | BCN · HA-ViD action segmentation, atomic action | F1: 20.9 -> 41.7 | Improved |
| atomic action level instead of primitive task level | action recognition · HA-ViD | Top-1: 15.6% lower on average | Hurt |
| right hand instead of left hand | atomic action recognition · HA-ViD | Top-1: 2.4% lower | Hurt |
| atomic action level instead of primitive task level | action segmentation · HA-ViD | Acc: 4% lower on average | Hurt |
| DINO Swin-L instead of Faster-RCNN ResNext101 | object detection · HA-ViD | AP: 22.2 -> 35.5 | Improved |
| ResNet101 backbone instead of ResNet50 | Faster-RCNN · HA-ViD | AP: 21.7 -> 20.9 | No clear effect |
| YOLOv5-large instead of YOLOv5-small | YOLOv5 · HA-ViD | AP: 10.2 -> 12.9 | Improved |
| ground truth boxes instead of DINO detections | SORT · HA-ViD MOT | MOTA: 20.4 -> 94.5 | Improved |
| ByteTrack instead of SORT with ground truth boxes | MOT tracker · HA-ViD MOT | IDS: 54.8 -> 121.6 | Hurt |
| ByteTrack instead of SORT with ground truth boxes | MOT tracker · HA-ViD MOT | MOTA: 94.5 -> 98.5 | Improved |
Closest papers by meaning
- Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activities (CVPR 2022 · similarity 0.88)
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.86)
- AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects (CVPR 2026 · similarity 0.86)
- From Observation to Action: Latent Action-based Primitive Segmentation for VLA Pre-training in Industrial Settings (CVPR 2026 · similarity 0.85)
- Aligning Step-by-Step Instructional Diagrams to Video Demonstrations (CVPR 2023 · similarity 0.85)
- IndEgo: A Dataset of Industrial Scenarios and Collaborative Work for Egocentric Assistants (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
