AI Paper Atlas

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

Highly Accurate Dichotomous Image Segmentation

Xuebin Qin, Hang Dai, Xiaobin Hu, Deng-Ping Fan, Ling Shao, Luc Van Gool

ECCV 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Amodal and open-world instance segmentation · 151 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
increase U2-Net input size from 320 to 512 (ablation)U2-Net · DIS-VDmaxF_beta: 0.748 -> 0.769Improved
increase U2-Net input size to 1024 without adaptation (ablation)U2-Net · DIS-VDmaxF_beta: 0.769 -> .764Mixed
add stride-2 input convolution (Adp) at 1024 input (ablation)U2-Net · DIS-VDmaxF_beta: .764 -> 0.776Improved
add intermediate supervision on last 6 decoder stages with L2 (ablation)IS-Net · DIS-VDmaxF_beta: 0.776 -> 0.790Improved
add intermediate supervision on last 6 decoder stages with L2 (ablation)IS-Net · DIS-VDHCE_gamma: 1076 -> 1056Improved
apply intermediate supervision only to last 1 decoder stage (ablation)IS-Net · DIS-VDmaxF_beta: 0.776 -> 0.777Improved
use L2 instead of KL for feature synchronization (ablation)IS-Net · DIS-VDmaxF_beta: 0.770 -> 0.790Improved
use L2 instead of L1 for feature synchronization (ablation)IS-Net · DIS-VDHCE_gamma: 1144 -> 1056Improved
use separate outconvs instead of sharing with GT encoder (ablation)IS-Net · DIS-VDmaxF_beta: 0.745 -> 0.790Improved
choose GT encoder seed 5289 over seed 1 (ablation)IS-Net · DIS-VDmaxF_beta: 0.786 -> 0.791No clear effect
choose GT encoder seed 5289 over seed 472 (ablation)IS-Net · DIS-VDHCE_gamma: 1071 -> 1116Hurt
use IS-Net instead of U2-NetIS-Net · DIS-VDHCE_gamma: 1413 -> 1116Improved

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.

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

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