AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

AttrSeg: Open-Vocabulary Semantic Segmentation via Attribute Decomposition-Aggregation

Chaofan Ma, Yuhuan Yang, Chen Ju, Fei Zhang, Ya Zhang, Yanfeng Wang

NeurIPS 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary semantic segmentation · 3 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use direct aggregation (all attributes in one sentence) instead of hierarchical (ablation)AttrSeg RN101 · PASCAL-5imIoU: 49.1 -> 44.8Hurt
Use pre-aggregation instead of hierarchical (ablation)AttrSeg RN101 · PASCAL-5imIoU: 49.1 -> 46.1Hurt
Use post-aggregation instead of hierarchical (ablation)AttrSeg RN101 · PASCAL-5imIoU: 49.1 -> 46.7Hurt
Remove cross attention on images in the clustering module (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 54.1Hurt
Remove cross attention on attributes (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 54.6Hurt
Remove both cross attentions (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 53.2Hurt
Remove the mixer (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 55.7Hurt
Remove cross attention and mixer (slot attention only) (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 52.9Hurt
Reduce aggregation stages from 4 to 2 with 15 attributes (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 52.4Hurt
Increase aggregation stages from 4 to 5 with 15 attributes (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 56.5No clear effect
Reduce number of input attributes from 15 to 5 (4 stages) (ablation)AttrSeg ViT-L · PASCAL-5imIoU: 56.4 -> 51.1Hurt
Add inaccurate attributes to clean attributes (ablation)AttrSeg · PASCAL-5imIoU: 59.5 -> 59.1No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy