Research map / Video / Action recognition
Large Scale Holistic Video Understanding
Ali Diba, Mohsen Fayyaz, Vivek Sharma, Manohar Paluri, Jürgen Gall, Rainer Stiefelhagen, Luc Van Gool
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 |
|---|---|---|---|
| add a 2D appearance branch to the 3D network (HATNet) | 3D-ResNet18 · HVU overall | mAP: 35.8 -> 40 | Improved |
| add the 2D appearance branch to the 3D-STCNet | 3D-STCNet · HVU overall | mAP: 36.7 -> 40 | Improved |
| train with six task-specific heads instead of single standard head | 3D-ResNet18 · HVU overall | mAP: 35.8 -> 37 | Improved |
| train with six task-specific heads instead of single standard head | HATNet · HVU overall | mAP: 40 -> 41.3 | Improved |
| pre-train on HVU instead of Kinetics | 3D-ResNet18 · UCF101 | mAP: 89.8 -> 90.5 | Improved |
| pre-train on HVU instead of Kinetics | 3D-ResNet18 · HMDB51 | mAP: 62.1 -> 65.1 | Improved |
| pre-train on HVU instead of training from scratch | 3D-ResNet18 · Kinetics | mAP: 65.6 -> 67.8 | Improved |
| add object tags to action-only training (ablation) | 3D-ResNet18 · HVU action category | accuracy: 43.6 -> 44.5 | Improved |
| add scene tags to action+object training (ablation) | 3D-ResNet18 · HVU action category | accuracy: 44.5 -> 45.6 | Improved |
| train all categories together instead of action only (ablation) | 3D-ResNet18 · HVU action category | accuracy: 43.6 -> 47.5 | Improved |
| pre-train on HVU instead of Kinetics for captioning | 3D-ResNet50+M · MSR-VTT | BLEU@4: 41.8 -> 42.7 | Improved |
| pre-train on HVU instead of Kinetics for clustering features | 3D-ResNet50 · ActivityNet-100 | clustering accuracy: 50.3 -> 53.5 | Improved |
Closest papers by meaning
- Harnessing Object and Scene Semantics for Large-Scale Video Understanding (CVPR 2016 · similarity 0.88)
- MLVU: Benchmarking Multi-task Long Video Understanding (CVPR 2025 · similarity 0.86)
- PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding (NeurIPS 2025 · similarity 0.86)
- Leader360V: A Large-scale, Real-world 360 Video Dataset for Multi-task Learning in Diverse Environment (NeurIPS 2025 · similarity 0.86)
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.86)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.85)
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.
