Research map / Detection and segmentation / CNN features
DynaMask: Dynamic Mask Selection for Instance Segmentation
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 |
|---|---|---|---|
| replace fixed-resolution Mask R-CNN masks with the DynaMask dual-level FPN and dynamic mask selection | Mask R-CNN, ResNet-50-FPN · COCO val2017, 1× | AP: 34.7 -> 37.6 | Helped |
| replace fixed-resolution Mask R-CNN masks with the DynaMask dual-level FPN and dynamic mask selection | Mask R-CNN, ResNet-101-FPN · COCO val2017, 1× | AP: 36.1 -> 38.7 | Helped |
| use DynaMask instead of Mask R-CNN | ResNet-50-FPN instance segmentation · Cityscapes val | AP: 33.8 -> 38.0 | Helped |
| use DynaMask instead of Mask R-CNN | R50-FPN instance segmentation · LVIS validation | AP: 22.1 -> 25.7 | Helped |
| increase the uniform mask resolution from 28×28 to 56×56 (ablation) | DynaMask · COCO val2017 | AP: 36.1 -> 37.1 | Helped |
| increase the uniform mask resolution from 56×56 to 112×112 (ablation) | DynaMask · COCO val2017 | AP: 37.1 -> 37.6 | Helped |
| increase the uniform mask resolution from 28×28 to 56×56 (ablation) | Mask R-CNN · COCO val2017 | AP: 34.7 -> 33.8 | Hurt |
| apply a budget constraint with Ct=1.0 and λ2=0.4 (ablation) | DynaMask · COCO val2017 | AP: 37.6 -> 37.6 | No clear effect |
| tighten the budget to Ct=0.4 and λ2=0.4 (ablation) | DynaMask · COCO val2017 | AP: 37.6 -> 36.8 | Hurt |
| select masks dynamically by instance segmentation difficulty rather than by object size (ablation) | DynaMask · COCO val2017 | AP: 35.9 -> 36.8 | Helped |
| replace standard FPN variants with DynaMask's adaptive dual-level FPN (ablation) | FPN instance segmentation · COCO val2017 | AP: 36.4 -> 37.6 | Helped |
Closest papers by meaning
- RefineMask: Towards High-Quality Instance Segmentation With Fine-Grained Features (CVPR 2021 · similarity 0.88)
- BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020 · similarity 0.88)
- ContrastMask: Contrastive Learning To Segment Every Thing (CVPR 2022 · similarity 0.87)
- SOLOv2: Dynamic and Fast Instance Segmentation (NeurIPS 2020 · similarity 0.87)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.87)
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.85)
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.