AI Paper Atlas

Research map / Detection and segmentation / CNN features

Holistically-Nested Edge Detection

Saining Xie, Zhuowen Tu

IJCV 2017 · Detection and segmentation › CNN features · approach: RGB-D salient object detection · 3,478 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 deep supervision on side outputs (ablation)HED fusion-output · BSDS500AP: .738 -> .787Improved
add deep supervision on side outputs (ablation)HED fusion-output · BSDS500ODS: .771 -> .782Improved
add deep supervision on side outputs (updated training) (ablation)HED fusion-output · BSDS500 (updated training)AP: .730 -> .811Improved
add deep supervision (updated training) (ablation)HED fusion-output · BSDS500 (updated training)ODS: .785 -> .790Improved
use HED topology instead of FCN-8sedge detection net · BSDS500ODS: .697 -> .782Improved
use HED topology instead of FCN-2s with extra low-level linksedge detection net · BSDS500ODS: .738 -> .782Improved
average side outputs 1-5 instead of learned fusion (ablation)HED · BSDS500AP: .787 -> .822Improved
merge fusion output with averaged side outputs (late merging)HED · BSDS500AP: .787 -> .833Improved
include side-output 5 in average of side outputs 1-4 (ablation)HED · BSDS500ODS: .760 -> .774Improved
add 100 extra training imagesHED · BSDS500ODS: .782 -> .797Improved
use RGB and HHA depth encoding instead of raw depth channel or concatenationHED · NYUDODS: .746Improved
retrain with scale augmentation and full-resolution imagesHED · BSDS500ODS: .782 -> .790Improved

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