Research map / Video / Video segmentation and tracking
BlockCopy: High-Resolution Video Processing With Block-Sparse Feature Propagation and Online Policies
Thomas Verelst, Tinne Tuytelaars
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 BlockCopy with tau=0.3 | CSP · CityPersons val | log-average miss rate, Reasonable (%): 11.0 -> 11.4 | Mixed |
| add BlockCopy with tau=0.7 | CSP · CityPersons val | log-average miss rate, Reasonable (%): 11.0 -> 11.1 | Mixed |
| add BlockCopy with tau=0.3 | CSP · CityPersons val | GMACs: 1128 -> 388 GMACs | Improved |
| add BlockCopy with tau=0.3 | Mask-RCNN · Cityscapes instance segmentation val | FPS: 6.7 -> 11.0 | Improved |
| add BlockCopy with tau=0.5 | Mask-RCNN · Cityscapes instance segmentation test | AP: 32.0 -> 31.7 | Mixed |
| add previous frame to policy input (ablation) | BlockCopy policy · CityPersons val | log-average miss rate, Reasonable (%): 13.0 -> 12.0 | Improved |
| add previous output to policy input (ablation) | BlockCopy policy · CityPersons val | log-average miss rate, Reasonable (%): 12.0 -> 11.7 | Improved |
| add previous execution grid to policy input (ablation) | BlockCopy policy · CityPersons val | log-average miss rate, Reasonable (%): 11.7 -> 11.4 | Improved |
| train policy online instead of offline (ablation) | BlockCopy policy · CityPersons val | log-average miss rate, Reasonable (%): 13.3 -> 11.4 | Improved |
| use ResNet-20 instead of ResNet-8 policy backbone (ablation) | BlockCopy policy · CityPersons val | log-average miss rate, Reasonable (%): 11.4 -> 11.5 | No clear effect |
| use BlockCopy instead of lower resolution and lower frame rate baselines | CSP · CityPersons val | log-average miss rate, Reasonable (%) | Improved |
Closest papers by meaning
- SBNet: Sparse Blocks Network for Fast Inference (CVPR 2018 · similarity 0.86)
- BlockDrop: Dynamic Inference Paths in Residual Networks (CVPR 2018 · similarity 0.86)
- Budget-Aware Deep Semantic Video Segmentation (CVPR 2017 · similarity 0.85)
- Block Image Compressive Sensing with Local and Global Information Interaction (AAAI 2024 · similarity 0.83)
- Compute Better Spent: Replacing Dense Layers with Structured Matrices (ICML 2024 · similarity 0.83)
- LiVOS: Light Video Object Segmentation with Gated Linear Matching (CVPR 2025 · 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.
