Research map / Detection and segmentation / Open-set and few-shot detection
FLOSS: Free Lunch in Open-vocabulary Semantic Segmentation
Yasser Benigmim, Mohammad Fahes, Tuan-Hung Vu, Andrei Bursuc, Raoul de Charette
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 |
|---|---|---|---|
| add FLOSS class-expert fusion | CLIP-DINOiser · Cityscapes | mIoU: 31.3 -> 34.6 | Improved |
| add FLOSS class-expert fusion | MaskCLIP · VOC20 | mIoU: 61.8 -> 61.8 | No clear effect |
| add FLOSS class-expert fusion | NACLIP · ADE20K | mIoU: 17.4 -> 18.4 | Improved |
| add FLOSS with experts identified on Cityscapes, evaluated on ACDC/BDD/Mapillary | CLIP-DINOiser · out-of-domain driving datasets (average) | mIoU: 26.4 -> 28.9 | Improved |
| add FLOSS with experts identified on Cityscapes | NACLIP · ACDC Fog | mIoU: 32.9 -> 32.5 | Hurt |
| identify experts from a single Cityscapes image | CLIP-DINOiser + FLOSS · out-of-domain driving datasets (average) | mIoU: 26.4 -> 28.3 | Improved |
| transfer experts identified on COCO-Stuff to other datasets | CLIP-DINOiser + FLOSS · Cityscapes (all classes) | mIoU: 31.1 -> 32.2 | Improved |
| fuse by highest-confidence expert instead of averaging all soft maps (ablation) | CLIP-DINOiser + FLOSS · Cityscapes | mIoU: 31.2 -> 34.6 | Improved |
| fuse by highest-confidence expert instead of falling back to averaged classifier on conflict (ablation) | CLIP-DINOiser + FLOSS · Cityscapes | mIoU: 32.0 -> 34.6 | Improved |
| fuse by highest-confidence expert instead of averaging conflicting experts (ablation) | CLIP-DINOiser + FLOSS · Cityscapes | mIoU: 34.4 -> 34.6 | Improved |
| use entropy instead of other unsupervised metrics to pick experts (ablation) | CLIP-DINOiser + FLOSS · Cityscapes | mIoU | Improved |
| use estimated experts instead of true-expert oracle at rho=100% (ablation) | CLIP-DINOiser + FLOSS · Cityscapes | mIoU: 34.1 -> 34.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CLIP-DINOiser — CLIP-DINOiser: Teaching CLIP a few DINO tricks for open-vocabulary semantic segmentation (ECCV 2024) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Open-Vocabulary Semantic Segmentation Models From Natural Language Supervision (CVPR 2023 · similarity 0.91)
- Auto-Vocabulary Semantic Segmentation (ICCV 2025 · similarity 0.90)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.89)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.89)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.89)
- Open-Vocabulary Semantic Segmentation with Image Embedding Balancing (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.
