Research map / Video / Video segmentation and tracking
Video Object Detection with an Aligned Spatial-Temporal Memory
Fanyi Xiao, Yong Jae Lee
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 MatchTrans memory alignment (ablation) | STMN · ImageNet VID test (VGG-16, ablation) | mAP: 49.0 -> 50.7 | Improved |
| use ReLU-based STMM instead of Sigmoid/Tanh ConvGRU, with pretrained FCs (ablation) | STMN · ImageNet VID test (VGG-16, ablation) | mAP: 48.0 -> 50.7 | Improved |
| initialize FC layers after the memory with pretrained weights instead of randomly (ablation) | ConvGRU · ImageNet VID test (VGG-16, ablation) | mAP: 44.8 -> 48.0 | Improved |
| increase test window size from 3 to 7 (ablation) | STMN · ImageNet VID val | mAP: -1.9%, 0.0% | Improved |
| increase test window size from 7 to 11 (ablation) | STMN · ImageNet VID val | mAP: 0.0%, +0.7% | Improved |
| increase test window size from 7 to 15 (ablation) | STMN · ImageNet VID val | mAP: 0.0%, +1.0% | Improved |
| add the spatial-temporal memory to the static detector | R-FCN ResNet-101 · ImageNet VID val | mAP: 73.4 -> 80.5 | Improved |
| add the spatial-temporal memory to Faster-RCNN VGG-16 | Fast-RCNN VGG-16 · ImageNet VID val | mAP: 52.2 -> 61.7 | Improved |
| replace FlowNet-style optical flow alignment with MatchTrans | STMN · alignment time per 337x600 frame | ms: 24.3ms -> 2.9ms | Improved |
Closest papers by meaning
- Learning Video Object Segmentation With Visual Memory (ICCV 2017 · similarity 0.89)
- You Don't Only Look Once: Constructing Spatial-Temporal Memory for Integrated 3D Object Detection and Tracking (ICCV 2021 · similarity 0.88)
- Alignment Before Aggregation: Trajectory Memory Retrieval Network for Video Object Segmentation (ICCV 2023 · similarity 0.88)
- Recurrent Dynamic Embedding for Video Object Segmentation (CVPR 2022 · similarity 0.88)
- Mobile Video Object Detection With Temporally-Aware Feature Maps (CVPR 2018 · similarity 0.88)
- LiVOS: Light Video Object Segmentation with Gated Linear Matching (CVPR 2025 · similarity 0.87)
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.
