Research map / Detection and segmentation / Open-set and few-shot detection
Bayesian Semantic Instance Segmentation in Open Set World
Trung Pham, Vijay B. G. Kumar, Thanh-Toan Do, Gustavo Carneiro, Ian Reid
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Bayesian global segmentation instead of thresholded-UCM baseline (no detections) | unsupervised segmentation · NYU 654 test images, unknown classes | F1@0.5 IoU: 47.8 -> 55.7 | Improved |
| Bayesian global segmentation instead of thresholded-UCM baseline (no detections) | unsupervised segmentation · NYU 654 test images, known classes | F1@0.5 IoU: 47.4 -> 45.6 | Mixed |
| add bounding box guidance | Our method · NYU 654 test images, known classes | F1@0.5 IoU: 45.6 -> 48.6 | Improved |
| add mask guidance | Our method · NYU 654 test images, known classes | F1@0.5 IoU: 45.6 -> 51.1 | Improved |
| add mask guidance | Our method · NYU 654 test images, unknown classes | F1@0.5 IoU: 55.7 -> 53.8 | Hurt |
| add mask guidance | Baseline (0.4) · NYU 654 test images, known classes | F1@0.5 IoU: 47.4 -> 7.3 | Hurt |
| add mask guidance | Baseline (0.3) · NYU 654 test images, known classes | F1@0.5 IoU: 40.1 -> 10.6 | Hurt |
| raise baseline UCM threshold from 0.3 to 0.4 (ablation) | Baseline · NYU 654 test images, known classes | F1@0.5 IoU: 40.1 -> 47.4 | Improved |
| Bayesian method with box supervision vs baseline with box supervision | Weak box supervision · COCO val 5k, 80 known classes | mAP: 10.1 -> 20.0 | Improved |
| Bayesian method with box supervision vs baseline with box supervision | Weak box supervision · COCO val 5k, 80 known classes | mIoU: 25.2 -> 32.3 | Improved |
| box-only supervision instead of full mask supervision | Mask-RCNN vs ours · COCO val 5k, 80 known classes | mAP: 30.5 -> 20.0 | Hurt |
| box-only supervision instead of full mask supervision | Mask-RCNN vs ours · COCO val 5k, 80 known classes | mIoU: 37.3 -> 32.3 | Hurt |
Closest papers by meaning
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · similarity 0.89)
- SOLO: Segmenting Objects by Locations (ECCV 2020 · similarity 0.89)
- Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity (CVPR 2022 · similarity 0.89)
- Open-World Semantic Segmentation Including Class Similarity (CVPR 2024 · similarity 0.88)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.88)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · 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.
