Research map / Detection and segmentation / CNN features
Recurrent Pixel Embedding for Instance Grouping
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 spherical pixel embeddings with recurrent mean-shift grouping | Object proposal generation · PASCAL VOC 2012 validation, 10 proposals | Average Recall at IoU=0.50: 0.558 -> 0.769 | Helped |
| Use spherical pixel embeddings with recurrent mean-shift grouping | Object proposal generation · PASCAL VOC 2012 validation, 60 proposals | Average Recall at IoU=0.50: 0.738 -> 0.814 | Helped |
| Add pixel-pair embedding loss alongside cross-entropy (ablation) | DeepLab-v3 semantic segmentation · PASCAL VOC 2012 validation | mIoU | Helped |
| Backpropagate training loss through recurrent mean-shift grouping (ablation) | Pixel embedding model · Pixel instance grouping | grouping quality | Helped |
| Accumulate loss over every mean-shift iteration rather than only the final iteration (ablation) | Gaussian Blurring Mean Shift grouping · Synthetic 1-D clustering toy example | cluster compactness | Helped |
| Add logistic-loss fine-tuning after boundary embedding training (ablation) | BSDS500 boundary detector · BSDS500 test | boundary representation | Helped |
Closest papers by meaning
- GroupViT: Semantic Segmentation Emerges From Text Supervision (CVPR 2022 · similarity 0.88)
- Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity (CVPR 2022 · similarity 0.88)
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · similarity 0.88)
- Instance Segmentation by Jointly Optimizing Spatial Embeddings and Clustering Bandwidth (CVPR 2019 · similarity 0.88)
- Tagger: Deep Unsupervised Perceptual Grouping (NeurIPS 2016 · similarity 0.87)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 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.