Research map / Video / Video segmentation and tracking
Temporal Context Enhanced Feature Aggregation for Video Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 temporal fusion (3 frames) to image-based R-FCN (ablation) | R-FCN · ImageNet VID val | mAP: 73.7 -> 76.0 | Helped |
| add two cascaded DeformAlign modules for feature alignment (ablation) | TCENet · ImageNet VID val | mAP: 76.0 -> 77.1 | Helped |
| add temporal and spatial attention (full TCEA) (ablation) | TCENet · ImageNet VID val | mAP: 77.1 -> 78.4 | Helped |
| add temporal stride predictor (ablation) | TCENet · ImageNet VID val | mAP: 78.4 -> 80.3 | Helped |
| add temporal stride predictor (ablation) | TCENet · ImageNet VID val | runtime (ms): 124 -> 125 | Hurt |
| use DeformAlign instead of flow-guided alignment (ablation) | TCENet · ImageNet VID val (DCN, SFEA, window 27) | mAP: 78.8 -> 79.8 | Helped |
| use DeformAlign instead of STSN alignment (ablation) | TCENet · ImageNet VID val (DCN, SFEA, window 27) | mAP: 78.9 -> 79.8 | Helped |
| shuffle temporal order of aggregated frames at test time (ablation) | TCENet · ImageNet VID val | mAP: 80.3 -> 14.3 | Hurt |
| reverse temporal order of aggregated frames at test time (ablation) | TCENet · ImageNet VID val | mAP: 80.3 -> 80.3 | No clear effect |
| shuffle frame order for SFEA aggregation (ablation) | SFEA · ImageNet VID val | mAP: 79.2 -> 79.2 | No clear effect |
| use TCEA rather than SFEA aggregation (ablation) | TCENet · ImageNet VID val | mAP: 79.2 -> 80.3 | Helped |
| fuse 7 frames instead of 3 at the same temporal receptive field (ablation) | lightweight TCENet · ImageNet VID val | mAP | Helped |
Closest papers by meaning
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.87)
- Temporal Context Aggregation Network for Temporal Action Proposal Refinement (CVPR 2021 · similarity 0.87)
- D2FANet: Enhancing Video Object Detection with Dual-Domain Feature Aggregation Network (CVPR 2026 · similarity 0.87)
- Feature Aggregated Queries for Transformer-Based Video Object Detectors (CVPR 2023 · similarity 0.86)
- Temporal ROI Align for Video Object Recognition (AAAI 2021 · similarity 0.85)
- Object Detection in Videos With Tubelet Proposal Networks (CVPR 2017 · 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.