Research map / Detection and segmentation / CNN features
Mask Encoding for Single Shot Instance Segmentation
Check 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 |
|---|---|---|---|
| Represent masks with a compact, explicitly encoded coefficient vector instead of directly predicting high-dimensional masks (ablation) | MEInst · COCO val2017 | mask AP: 29.7 -> 31.8 | Helped |
| Increase mask-encoding dimensions from 20 to 60 (ablation) | MEInst · COCO val2017 | mask AP: 29.8 -> 31.8 | Helped |
| Increase mask-encoding dimensions from 60 to 80 (ablation) | MEInst · COCO val2017 | mask AP: 31.8 -> 31.9 | No clear effect |
| Use l2 rather than cosine similarity as the mask loss (ablation) | MEInst · COCO val2017 | mask AP: 28.9 -> 31.8 | Helped |
| Use a large-kernel prediction layer (ablation) | MEInst · COCO val2017 | mask AP: +0.7% AP | Helped |
| Use deformable convolution in the last head layer (ablation) | MEInst · COCO val2017 | mask AP: +1.5% improvement in AP | Helped |
| Add a mask prediction branch and train with mask supervision (ablation) | FCOS · COCO val2017 | box AP: 39.6 -> 40.4 | Helped |
| Use MEInst rather than ESE-Seg for mask prediction | ResNet-50 instance segmentation · COCO val2017 | mask AP: 21.6 -> 30.3 | Helped |
| Use MEInst rather than Mask R-CNN for small objects | COCO instance segmentation · COCO test-dev, small objects | mask AP: 16.9 -> 21.8 | Helped |
| Use MEInst rather than Mask R-CNN for large objects | COCO instance segmentation · COCO test-dev, large objects | mask AP: 53.5 -> 46.7 | Hurt |
Closest papers by meaning
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.90)
- BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020 · similarity 0.90)
- SOLOv2: Dynamic and Fast Instance Segmentation (NeurIPS 2020 · similarity 0.90)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.89)
- DCT-Mask: Discrete Cosine Transform Mask Representation for Instance Segmentation (CVPR 2021 · similarity 0.89)
- SIM: Semantic-Aware Instance Mask Generation for Box-Supervised Instance Segmentation (CVPR 2023 · similarity 0.89)
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.