Research map / Video / Video segmentation and tracking
Learning Where to Focus for Efficient Video Object Detection
Zhengkai Jiang, Yu Liu, Ceyuan Yang, Jihao Liu, Peng Gao, Qian Zhang, Shiming Xiang, Chunhong Pan
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 learning of sampled locations (Gaussian init) (ablation) | LSTS · ImageNet VID val | mAP (%): 75.5 -> 77.2 | Improved |
| add learning of sampled locations (Uniform init) (ablation) | LSTS · ImageNet VID val | mAP (%): 75.5 -> 76.8 | Improved |
| use Gaussian instead of Uniform initialization of sampled locations (ablation) | LSTS · ImageNet VID val | mAP (%): 76.8 -> 77.2 | Improved |
| use LSTS instead of Non-Local for feature propagation (ablation) | SRFU+DFA framework · ImageNet VID val | mAP (%): 74.2 -> 77.2 | Improved |
| use LSTS instead of MatchTrans for feature propagation (ablation) | SRFU+DFA framework · ImageNet VID val | mAP (%): 75.5 -> 77.2 | Improved |
| use LSTS instead of optical flow warping (full model) (ablation) | SRFU+DFA framework · ImageNet VID val | mAP (%): 76.1 -> 77.2 | Improved |
| use LSTS instead of optical flow warping (full model) (ablation) | SRFU+DFA framework · ImageNet VID val | parameters (M): 100.5 -> 64.5 | Improved |
| use LSTS instead of optical flow warping (full model) (ablation) | SRFU+DFA framework · ImageNet VID val | runtime (FPS): 18.9 -> 23.0 | Improved |
| add quality-aware memory update to keyframes (ablation) | SRFU · ImageNet VID val | mAP (%): 73.5 -> 75.9 | Improved |
| quality-aware instead of fixed 0.5 weights for keyframe memory update (ablation) | SRFU · ImageNet VID val | mAP (%): 75.8 -> 75.9 | Improved |
| add non-keyframe aggregation (DFA) (ablation) | DFA · ImageNet VID val | mAP (%): 75.9 -> 76.4 | Improved |
| add non-keyframe transform unit (ablation) | DFA · ImageNet VID val | mAP (%): 76.4 -> 76.5 | Improved |
Closest papers by meaning
- Object Detection in Video with Spatiotemporal Sampling Networks (ECCV 2018 · similarity 0.86)
- Video Object Detection via Object-level Temporal Aggregation (ECCV 2020 · similarity 0.86)
- Progressive Sparse Local Attention for Video Object Detection (ICCV 2019 · similarity 0.86)
- Flow-Guided Feature Aggregation for Video Object Detection (ICCV 2017 · similarity 0.85)
- Scale-Aware Spatio-Temporal Relation Learning for Video Anomaly Detection (ECCV 2022 · similarity 0.85)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.84)
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.
