Research map / Video / Video segmentation and tracking
TDViT: Temporal Dilated Video Transformer for Dense Video Tasks
Guanxiong Sun, Yang Hua, Guosheng Hu, Neil Robertson
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 |
|---|---|---|---|
| replace Swin-T backbone with TDViT-T | Faster R-CNN · ImageNet VID | AP box: 47.1 -> 49.1 | Improved |
| replace Swin-S backbone with TDViT-S | Faster R-CNN · ImageNet VID | AP box: 52.6 -> 55.4 | Improved |
| add advanced variant (two extra TDTBs in stage 3) to TDViT-T | TDViT-T · ImageNet VID | AP box: 49.1 -> 50.9 | Improved |
| replace R-101 backbone with TDViT-T in SELSA | SELSA · ImageNet VID | AP50 box: 81.5 -> 83.9 | Improved |
| replace Swin-T with TDViT-T | MaskTrack R-CNN · YouTube-VIS 2019 | AP mask: 33.7 -> 35.4 | Improved |
| replace Swin-S with TDViT-S | MaskTrack R-CNN · YouTube-VIS 2019 | AP mask: 36.8 -> 37.7 | Improved |
| use split scheme rather than factorised scheme (ablation) | TDViT-T · ImageNet VID | AP50 box: 78.0 -> 78.5 | Improved |
| add 2 more TDTBs in stage 3 (6 to 8 blocks) (ablation) | TDViT-T · ImageNet VID | AP50 box: 78.5 -> 79.9 | Improved |
| increase temporal dilations from (1,2,4,8) to (4,8,16,32) (ablation) | TDViT-T · ImageNet VID | AP50 box: 77.2 -> 78.5 | Improved |
| increase temporal dilations further to (5,10,20,40) (ablation) | TDViT-T · ImageNet VID | AP50 box: 78.5 -> 78.4 | No clear effect |
| use patch shuffle sampling instead of temporal earliest (ablation) | TDViT-T · ImageNet VID | AP50 box: 78.5 -> 78.8; FPS 23.9 -> 19.7 | Mixed |
| use channel shuffle sampling instead of temporal earliest (ablation) | TDViT-T · ImageNet VID | AP50 box: 78.5 -> 77.7 | Hurt |
Closest papers by meaning
- Temporally Efficient Vision Transformer for Video Instance Segmentation (CVPR 2022 · similarity 0.89)
- TempFormer: Temporally Consistent Transformer for Video Denoising (ECCV 2022 · similarity 0.89)
- ViViT: A Video Vision Transformer (ICCV 2021 · similarity 0.87)
- TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection (AAAI 2025 · similarity 0.84)
- Adapting Short-Term Transformers for Action Detection in Untrimmed Videos (CVPR 2024 · similarity 0.84)
- Memory Consolidation Enables Long-Context Video Understanding (ICML 2024 · similarity 0.84)
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.
