Research map / Detection and segmentation / CNN features
Deep Variational 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 |
|---|---|---|---|
| Use DVIS with DeepLab-v3 (Xception 65) instead of the ResNet-50-FCN configuration | DVIS · PASCAL VOC 2012 val | AP average: 54.5 -> 56.6 | Helped |
| Use a large 128-pixel range for permutation-invariant pixel pairs rather than a local range (ablation) | DVIS · PASCAL VOC 2012 val | AP average: 44.59 -> 56.6 (range 8 -> range 128) | Helped |
| Add Mumford-Shah regularization (ablation) | DVIS · PASCAL VOC 2012 val | contour F1 at distance tolerance 1: 21.6 -> 23.5 | Helped |
| Add quantization loss alongside Mumford-Shah regularization (ablation) | DVIS · PASCAL VOC 2012 val | contour F1 at distance tolerance 10: 69.9 -> 71.7 | Helped |
| Increase the IoU score's weight in the verification confidence (ablation) | DVIS classification and verification · PASCAL VOC 2012 val | AP at high IoU thresholds | Helped |
| Use search-free global prediction instead of proposal-based search | DVIS · COCO test-dev 2017 | FPS: 38.0 FPS (DVIS ResNet-50); 28.7 FPS (YOLACT-700 ResNet-101) | Helped |
| Use DVIS instead of search-based methods on small objects | DVIS · COCO test-dev 2017 | AP_S: 11.0 (ResNet-50); 12.5 (ResNet-101); 12.1 (YOLACT-700) | Hurt |
Closest papers by meaning
- Fully Convolutional Instance-Aware Semantic Segmentation (CVPR 2017 · similarity 0.89)
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · similarity 0.88)
- Instance-Aware Semantic Segmentation via Multi-Task Network Cascades (CVPR 2016 · similarity 0.87)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.87)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · similarity 0.86)
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · similarity 0.86)
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.