Research map / Detection and segmentation / CNN features
K-Net: Towards Unified Image 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 |
|---|---|---|---|
| K-Net vs Panoptic FCN | panoptic segmentation · COCO panoptic val | PQ: 43.6 -> 47.1 | Helped |
| K-Net-N256 vs Cascade Mask R-CNN | instance segmentation · COCO val2017 | FPS: 10.3 -> 19.8 | Helped |
| add K-Net dynamic kernels to FCN | FCN · ADE20K | mIoU: 36.7 -> 43.3 | Helped |
| add K-Net to UperNet | UperNet Swin-L · ADE20K | mIoU: 50.6 -> 52 | Helped |
| add adaptive kernel update (ablation) | K-Net · COCO val2017 | mask AP: 31.2 -> 34.1 | Helped |
| add kernel interaction (ablation) | K-Net · COCO val2017 | mask AP: 22.6 -> 34.1 | Helped |
| add positional encoding (ablation) | K-Net · COCO val2017 | mask AP: 30.9 -> 34.1 | Helped |
| add coordconv on top of positional encoding (ablation) | K-Net · COCO val2017 | mask AP: 34.1 -> 34.0 | No clear effect |
| increase update rounds 1 -> 3 (ablation) | K-Net · COCO val2017 | mask AP: 21.8 -> 34.1 | Helped |
| increase update rounds beyond 3 (ablation) | K-Net · COCO val2017 | mask AP: 34.1 -> 34.5 | No clear effect |
| increase update rounds beyond 5 (ablation) | FCN + K-Net · ADE20K | mIoU: 44.1 -> 42.6 | Hurt |
| increase kernels 100 -> 256 (ablation) | K-Net · COCO val2017 | mask AP: 34.1 -> 34.7 | Helped |
Closest papers by meaning
- Video K-Net: A Simple, Strong, and Unified Baseline for Video Segmentation (CVPR 2022 · similarity 0.93)
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · similarity 0.90)
- Panoptic Feature Pyramid Networks (CVPR 2019 · similarity 0.90)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.88)
- COCONut: Modernizing COCO Segmentation (CVPR 2024 · similarity 0.88)
- Unsupervised Universal Image Segmentation (CVPR 2024 · 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.