Research map / Video / Video segmentation and tracking
Fast Video Object Segmentation via Dynamic Targeting Network
Lu Zhang, Zhe Lin, Jianming Zhang, Huchuan Lu, You He
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 finetuning on DAVIS training set (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean: 66.8 -> 83.6 | Improved |
| add pre-training on instance segmentation (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean: 69.2 -> 83.6 | Improved |
| add dynamic targeting network (RoI Align zoom-in) (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean: 72.3 -> 83.6 | Improved |
| use conditional RPN instead of original RPN (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean: 81.6 -> 83.6 | Improved |
| add mask guidance from warped previous mask in MRN (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean: 81.9 -> 83.6 | Improved |
| decision module versus random assignment of frames to box propagation (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean at 14.0 fps: 83.6 (decision module) vs 83.6 random at (83.6,14.0) | Improved |
| use box re-identification only instead of adaptive switching (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean / fps: 84.6 G Mean at 9.1 fps vs 83.6 at 14.0 fps | Mixed |
| use box propagation only instead of adaptive switching (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean / fps: 76.0 G Mean at 23.1 fps vs 83.6 at 14.0 fps | Mixed |
| increase decision threshold theta_c (ablation) | Dynamic Targeting Network · DAVIS-2016 val | G Mean vs fps | Mixed |
| compare with RGMP without online finetuning | Dynamic Targeting Network · DAVIS-2016 val | J Mean: 81.5 -> 83.7 | Improved |
| compare with RGMP on DAVIS-2017 | Dynamic Targeting Network · DAVIS-2017 val | J Mean: 64.8 -> 64.2 | Hurt |
Closest papers by meaning
- Learning Video Object Segmentation From Static Images (CVPR 2017 · similarity 0.89)
- Fast Video Object Segmentation by Reference-Guided Mask Propagation (CVPR 2018 · similarity 0.88)
- Motion-Guided Cascaded Refinement Network for Video Object Segmentation (CVPR 2018 · similarity 0.88)
- Learning Video Object Segmentation With Visual Memory (ICCV 2017 · similarity 0.88)
- Efficient Video Object Segmentation via Network Modulation (CVPR 2018 · similarity 0.88)
- Learning to Segment Moving Objects (IJCV 2018 · similarity 0.88)
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.
