AI Paper Atlas

Research map / Detection and segmentation / CNN features

Region-aware Distribution Contrast: A Novel Approach to Multi-Task Partially Supervised Learning

Meixuan Li, Tianyu Li, Guoqing Wang, Peng Wang, Yang Yang, Jie Zou

ECCV 2024 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 1 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 SAM regions instead of patches for local contrast (ablation)Gaussian contrast on HRNet18 · NYU-v2 one labelNormals mErr: 7.7141 -> 4.8613Improved
use SAM regions instead of patches for local contrast (ablation)Gaussian contrast on HRNet18 · NYU-v2 one labelSemantic mIoU: 41.56 -> 42.28Improved
model regions as Gaussian distributions instead of vectors (ablation)Region-aware contrast · NYU-v2 one labelNormals mErr: 5.8098 -> 4.8613Improved
model regions as Gaussian distributions instead of vectors (ablation)Region-aware contrast · Cityscapes one labelDepth aErr: 0.0145 -> 0.0121Improved
Gaussian distribution instead of pixel contrast (ablation)Region-aware contrast · NYU-v2 random labelSemantic mIoU: 46.41 -> 46.21Mixed
use Wasserstein distance instead of KL divergence (ablation)Gaussian region contrast · NYU-v2 random labelNormals mErr: 5.04 -> 4.49Improved
use Wasserstein distance instead of KL divergence (ablation)Gaussian region contrast · NYU-v2 one labelSemantic mIoU: 41.15 -> 42.28Improved
add region-wise Gaussian contrast to TaskExpertTaskExpert ViT-Large · NYU-v2 one labelSemantic mIoU: 49.53 -> 55.81Improved
add region-wise Gaussian contrast in the fully supervised settingMTL HRNet18 · NYU-v2 fully supervisedNormals mErr: 6.7298 -> 3.2878Improved
use SAM regions with vector contrast only (ablation)Region-aware contrast · NYU-v2 fully supervisedSemantic mIoU: 47.20 -> 47.64No 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