Research map / Detection and segmentation / CNN features
TransVOD: End-to-End Video Object Detection With Spatial-Temporal Transformers
Qianyu Zhou, Xiangtai Li, Lu He, Yibo Yang, Guangliang Cheng, Yunhai Tong, Lizhuang Ma, Dacheng Tao
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 TDTE and TDTD to single-frame baseline (ablation) | Deformable DETR baseline · ImageNet VID val (ResNet-50) | mAP (%): 76.0 -> 77.1 | Improved |
| add TQE to single-frame baseline (ablation) | Deformable DETR baseline · ImageNet VID val (ResNet-50) | mAP (%): 76.0 -> 78.9 | Improved |
| add TDTD to TQE (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 78.9 -> 79.3 | Improved |
| add TDTE to TQE+TDTD (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 79.3 -> 79.9 | Improved |
| add TDTE to TQE+TDTD with stronger Swin-Base backbone (ablation) | TransVOD · ImageNet VID val (Swin-Base) | mAP (%): 89.6 -> 89.6 | No clear effect |
| use more than 1 TDTE encoder layer (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 77.7 -> 77.6 | No clear effect |
| increase TDTD decoder layers from 1 to 6 (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 78.2 -> 72.3 | Hurt |
| increase TQE encoder layers from 1 to 3 (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 78.8 -> 79.6 | Improved |
| increase reference frames from 2 to 14 (ablation) | TransVOD · ImageNet VID val (ResNet-50) | mAP (%): 77.7 -> 79.3 | Improved |
| use coarse-to-fine query selection (80,50,20) instead of coarse-to-coarse (30,30,30) (ablation) | TQE · ImageNet VID val (ResNet-50) | mAP (%): 79.3 -> 79.9 | Improved |
| add RoI and Query Fusion (ablation) | TransVOD++ · ImageNet VID val (Swin-Base) | mAP50 (%): 88.3 -> 89.7 | Improved |
| add Hard Query Mining (ablation) | TransVOD++ · ImageNet VID val (Swin-Base) | mAP50 (%): 89.7 -> 90.0 | Improved |
Later papers that compared with this method
5 later accepted papers report TransVOD++ as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- MSTDiff: Multiscale-Aware Transformer Diffusion Network for Video Object Detection (AAAI 2026 · better on 2 of 2 shared results)
- D2FANet: Enhancing Video Object Detection with Dual-Domain Feature Aggregation Network (CVPR 2026 · better on 1 of 1 shared result)
- YOLOV: Making Still Image Object Detectors Great at Video Object Detection (AAAI 2023 · better on 1 of 1 shared result)
- Identity-Consistent Aggregation for Video Object Detection (ICCV 2023 · better on 1 of 1 shared result)
- PTSEFormer: Progressive Temporal-Spatial Enhanced TransFormer towards Video Object Detection (ECCV 2022 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.91)
- Temporal Object-Aware Vision Transformer for Few-Shot Video Object Detection (AAAI 2026 · similarity 0.90)
- TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection (AAAI 2025 · similarity 0.89)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.89)
- QDETRv: Query-Guided DETR for One-Shot Object Localization in Videos (AAAI 2024 · similarity 0.88)
- KD-DETR: Knowledge Distillation for Detection Transformer with Consistent Distillation Points Sampling (CVPR 2024 · similarity 0.87)
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.
