Research map / Video / Video segmentation and tracking
Temporal ROI Align for Video Object Recognition
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 ROI Align to SELSA-ReIm | SELSA-ReIm with R101 · ImageNet VID validation | mAP@IoU=0.5 (%): 80.3 -> 82.0 | Helped |
| Add Temporal ROI Align to SELSA-ReIm | SELSA-ReIm with X101 · ImageNet VID validation | mAP@IoU=0.5 (%): 83.0 -> 84.3 | Helped |
| Add Temporal ROI Align to SELSA-ReIm | SELSA-ReIm · EPIC KITCHENS S1 | mAP@0.5: 38.8 -> 42.2 | Helped |
| Add Temporal ROI Align to SELSA-ReIm | SELSA-ReIm · EPIC KITCHENS S2 | mAP@0.5: 36.7 -> 39.6 | Helped |
| Add Temporal ROI Align to MaskTrack R-CNN | MaskTrack R-CNN with R50-FPN · YouTube-VIS validation | AP: 30.3 -> 33.5 | Helped |
| Add Temporal ROI Align to MaskTrack R-CNN | MaskTrack R-CNN with X101-FPN · YouTube-VIS validation | AP: 34.9 -> 38.0 | Helped |
| Replace simple averaging with TAFA (ablation) | MS ROI Align feature aggregation · ImageNet VID | mAP (%): 78.5 -> 80.5 | Helped |
| Use MS ROI Align rather than Modulated Deformable ROI Align (ablation) | Temporal ROI Align with TAFA · ImageNet VID | mAP (%): 76.6 -> 80.5 | Helped |
| Use similarity-based ROI feature extraction instead of ROI Align plus Non-Local | Temporal ROI Align · ImageNet VID | mAP (%): 78.4 -> 80.5 | Helped |
| Increase MS ROI Align top-K locations from 1 to 4 (ablation) | Temporal ROI Align · ImageNet VID | mAP (%): 79.3 -> 80.5 | Helped |
| Increase MS ROI Align top-K locations beyond 4 (ablation) | Temporal ROI Align · ImageNet VID | mAP (%): 80.5 -> 80.4; 80.5 -> 80.2 | Hurt |
| Increase temporal attention blocks from 1 to 4 (ablation) | TAFA · ImageNet VID | mAP (%): 79.5 -> 80.5 | Helped |
Closest papers by meaning
- Semantic-Aware Transformation-Invariant RoI Align (AAAI 2024 · similarity 0.86)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.84)
- Anchor-Aware Similarity Cohesion in Target Frames Enables Predicting Temporal Moment Boundaries in 2D (CVPR 2025 · similarity 0.84)
- Don't Judge by the Look: Towards Motion Coherent Video Representation (ICLR 2024 · similarity 0.82)
- TE-TAD: Towards Full End-to-End Temporal Action Detection via Time-Aligned Coordinate Expression (CVPR 2024 · similarity 0.82)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · 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.