Research map / Detection and segmentation / CNN features
Towards High Performance 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 |
|---|---|---|---|
| Use sparse recursive feature aggregation instead of sparse feature propagation alone | Video object detection on ImageNet VID · ImageNet VID validation | mAP: almost 1% | Helped |
| Use sparse recursive feature aggregation rather than dense feature aggregation | Video object detection · ImageNet VID validation | runtime: 10 × speedup with 2% accuracy loss | Helped |
| Add spatially-adaptive partial feature updating | Ours (c1) · ImageNet VID validation | mAP: almost 2% | Helped |
| Add spatially-adaptive partial feature updating | Ours (c1) · ImageNet VID validation | mAP: 1.8% | Helped |
| Add spatially-adaptive partial feature updating | Ours (c1) · ImageNet VID validation | mAP: 1.4% | Helped |
| Use temporally-adaptive rather than fixed key-frame scheduling | Ours (c2) · ImageNet VID validation | mAP: 0.5% ∼ 1% | Helped |
| Use temporally-adaptive key-frame scheduling | Ours (c2) · ImageNet VID validation | mAP: nearly 2% | Helped |
| Replace fixed scheduling with oracle key-frame scheduling (ablation) | Proposed method · ImageNet VID validation | mAP: 80.9% mAP score at 22.8 fps runtime speed | Helped |
| Use recursive rather than non-recursive feature aggregation on sparse key frames (ablation) | Ours (c1) · ImageNet VID validation | mAP | Helped |
| Vary key-frame duration; use l = 10 frames (ablation) | Ours (c2) · ImageNet VID validation | speed-accuracy trade-off | Helped |
| Use small rather than tuned key-frame duration (ablation) | Ours (c2) · ImageNet VID validation | mAP | Hurt |
| Use large rather than tuned key-frame duration (ablation) | Ours (c2) · ImageNet VID validation | mAP | Hurt |
Closest papers by meaning
- YOLOV: Making Still Image Object Detectors Great at Video Object Detection (AAAI 2023 · similarity 0.86)
- Optimizing Video Object Detection via a Scale-Time Lattice (CVPR 2018 · similarity 0.85)
- Feature Aggregated Queries for Transformer-Based Video Object Detectors (CVPR 2023 · similarity 0.85)
- Minimum Delay Moving Object Detection (CVPR 2017 · similarity 0.85)
- Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (CVPR 2017 · similarity 0.84)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.83)
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.