AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

Max Pooling with Vision Transformers Reconciles Class and Shape in Weakly Supervised Semantic Segmentation

Simone Rossetti, Damiano Zappia, Marta Sanzari, Marco Schaerf, Fiora Pirri

ECCV 2022 · Detection and segmentation › Weakly supervised detection and segmentation · approach: CAM refinement and prototypes · 83 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
add equivariance loss L_ET with two branches (ViT-S/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 44.0 -> 59.2 (+15.2)Improved
add equivariance loss L_ET with two branches (ViT-S/16) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 43.3 -> 56.4 (+13.1)Improved
add equivariance loss L_ET with two branches (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 45.6 -> 65.1 (+19.5)Improved
add equivariance loss L_ET with two branches (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 44.1 -> 62.4 (+18.3)Improved
add HV-BiLSTM patch conditioning (ViT-S/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 59.2 -> 63.6 (+4.4)Improved
add HV-BiLSTM patch conditioning (ViT-S/16) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 56.4 -> 61.8 (+5.4)Improved
add HV-BiLSTM patch conditioning (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 65.1 -> 67.7 (+2.6)Improved
add HV-BiLSTM patch conditioning (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 62.4 -> 66.0 (+3.6)Improved
add dense CRF post-processing (ViT-S/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 63.6 -> 67.1 (+3.5)Improved
add dense CRF post-processing (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 trainmIoU: 67.7 -> 71.4 (+3.7)Improved
add dense CRF post-processing (ViT-B/16) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 66.0 -> 69.3 (+3.3)Improved
larger ViT backbone, ViT-S/16 to ViT-B/16 (full model with CRF) (ablation)ViT-PCM · PASCAL VOC 2012 valmIoU: 64.9 -> 69.3Improved

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