Research map / Detection and segmentation / CNN features
Memory Enhanced Global-Local Aggregation for Video Object Detection
Check 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 |
|---|---|---|---|
| combine global and local aggregation with long-range memory | MEGA · ImageNet VID validation | mAP: 81.4 vs 82.9 | Helped |
| replace LRM with a larger local span (ablation) | MEGA base model · ImageNet VID validation | mAP: 81.9 vs 82.9 | Hurt |
| add global feature aggregation (ablation) | MEGA · ImageNet VID validation | mAP: 81.3 vs 82.9 | Helped |
| add local feature aggregation (ablation) | MEGA · ImageNet VID validation | mAP: 81.8 vs 82.9 | Helped |
| increase global reference frames from 10 to 20 (ablation) | MEGA · ImageNet VID validation | mAP: 82.9 -> 83.0 | No clear effect |
| increase local reference frames from 25 to 29 (ablation) | MEGA · ImageNet VID validation | mAP: 82.9 -> 82.9 | No clear effect |
| reduce local relation modules from three to one (ablation) | MEGA · ImageNet VID validation | mAP: 82.9 vs 82.1 | Hurt |
| increase memory size from 25 to 45 (ablation) | MEGA · ImageNet VID validation | mAP: 82.9 -> 83.0 | No clear effect |
| include location information in global relation weights (ablation) | MEGA · ImageNet VID validation | mAP: 82.5 vs 82.9 | Hurt |
| restrict aggregation to current and previous frames | MEGA · ImageNet VID validation, online | mAP: 80.0 vs 81.9 | Helped |
Closest papers by meaning
- Efficient Regional Memory Network for Video Object Segmentation (CVPR 2021 · similarity 0.88)
- MAMBA: Multi-level Aggregation via Memory Bank for Video Object Detection (AAAI 2021 · similarity 0.87)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.87)
- Re-thinking Temporal Search for Long-Form Video Understanding (CVPR 2025 · similarity 0.86)
- Learning Object State Changes in Videos: An Open-World Perspective (CVPR 2024 · similarity 0.85)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · 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.