AI Paper Atlas

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

Multi-label Affordance Mapping from Egocentric Vision

Lorenzo Mur-Labadia, Jose J. Guerrero, Ruben Martinez-Cantin

ICCV 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Object-level representation learning · 21 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
replace multi-class softmax with independent binary classifiers trained with Asymmetric loss (ablation)DeepLab-v3 · easy-EPIC Aff testmIoU (%): 19.5 -> 42.3Improved
replace multi-class softmax with independent binary classifiers trained with Asymmetric loss (ablation)FPN · easy-EPIC Aff testmIoU (%): 12.9 -> 39.8Improved
use Asymmetric loss instead of plain BCE (ablation)DeepLab-v3 · easy-EPIC Aff testmIoU (%): 31.3 -> 42.3Improved
use Asymmetric loss instead of plain BCE (ablation)FPN · easy-EPIC Aff testmIoU (%): 22.2 -> 39.8Improved
use Asymmetric loss instead of plain BCE (ablation)U-Net · easy-EPIC Aff testmIoU (%): 20.9 -> 17.7Hurt
use dynamic threshold instead of top-K selection for multi-class heuristics (ablation)multi-class segmentation models · easy-EPIC Aff testmIoU (%): 8.9 -> 12.9 (FPN)Improved
increase number of winning classes in top-K heuristic (ablation)multi-class segmentation models · easy-EPIC Aff testmIoUHurt
use DeepLab-v3 decoder instead of FPN or U-Net (ablation)Asym multi-label segmentation · easy-EPIC Aff testmIoU (%): 17.7 (U-Net) / 39.8 (FPN) -> 42.3Improved
use Mask R-CNN instead of best multi-label segmentation modeldetection metrics · easy-EPIC Aff testmAP (%): Mask-RCNN 59.3 mAP, 62.6 AP50Improved
increase affordance classes from 20 to 43 (ablation)DeepLab-v3 Asym · EPIC-AffmIoU (%): 42.3 -> 31.1Hurt

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