Research map / Detection and segmentation / Open-set and few-shot detection
Toyota Smarthome Untrimmed: Real-World Untrimmed Videos for Activity Detection
Rui Dai, Srijan Das, Saurav Sharma, Luca Minciullo, Lorenzo Garattoni, François Brémond, Gianpiero Francesca
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 3D-pose attention module to SD-TCN on RGB (AGNet) (ablation) | I3D+SD-TCN · TSU fine-grained (cross-subject) | per-frame mAP: 29.2 -> 33.2 | Improved |
| add 3D-pose attention module to SD-TCN on RGB (AGNet) (ablation) | I3D+SD-TCN · TSU fine-grained (cross-view) | per-frame mAP: 18.3 -> 23.2 | Improved |
| use I3D instead of Inception RGB features (ablation) | SD-TCN · TSU fine-grained (cross-subject) | per-frame mAP: 22.3 -> 29.2 | Improved |
| use 3D pose (AGCN) instead of RGB (I3D) features (ablation) | SD-TCN · TSU fine-grained (cross-view) | per-frame mAP: 18.3 -> 22.4 | Improved |
| use 3D pose (AGCN) instead of RGB (I3D) features (ablation) | SD-TCN · TSU fine-grained (cross-subject) | per-frame mAP: 29.2 -> 26.2 | Hurt |
| add SD-TCN temporal modelling to I3D bottleneck | I3D+Bottleneck · TSU fine-grained (cross-subject) | per-frame mAP: 15.7 -> 29.2 | Improved |
| train I3D on trimmed clips instead of random untrimmed snippets (ablation) | I3D+Bottleneck · TSU fine-grained (cross-subject) | per-frame mAP: 15.7 -> 7.4 | Hurt |
| use balanced annotation instead of fine-grained annotation | I3D+SD-TCN · Balanced TSU (cross-subject) | per-frame mAP: 29.2 -> 35.8 | Improved |
| use balanced annotation instead of fine-grained annotation | I3D+Bottleneck · Balanced TSU (cross-subject) | per-frame mAP: 15.7 -> 17.1 | Improved |
| train with synchronized joint views (late fusion) instead of mixed view | SD-TCN · TSU joint-view (cross-subject) | per-frame mAP: 35.5 -> 38.9 | Improved |
| train with synchronized joint views using AG-Net instead of mixed view | AG-Net · TSU joint-view (cross-subject) | per-frame mAP: 35.5 -> 40.1 | Improved |
| use SSD human crops instead of center crops (ablation) | I3D+AGNet · TSU-CS | per-frame mAP: 26.9 -> 33.2 | Improved |
Closest papers by meaning
- Toyota Smarthome: Real-World Activities of Daily Living (ICCV 2019 · similarity 0.91)
- Deep Learning-Based Action Detection in Untrimmed Videos: A Survey (TPAMI 2022 · similarity 0.85)
- Efficient Activity Detection in Untrimmed Video with Max-Subgraph Search (TPAMI 2016 · similarity 0.84)
- Drive&Act: A Multi-Modal Dataset for Fine-Grained Driver Behavior Recognition in Autonomous Vehicles (ICCV 2019 · similarity 0.84)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.83)
- UniMTS: Unified Pre-training for Motion Time Series (NeurIPS 2024 · similarity 0.82)
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.
