AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection

Jing Zhang, Jianwen Xie, Nick Barnes

ECCV 2020 · Label-efficient and robust learning › Noisy labels · approach: Contrastive learning on noisy data · 64 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 noise generator network (f1 & l_s -> full) (ablation)Noise-aware encoder-decoder · DUTSS-measure: .668 -> .828Improved
add noise generator network (ablation)Noise-aware encoder-decoder · DUTSF-measure: .519 -> .747Improved
add edge-aware smoothness loss to f1 (ablation)f1 encoder-decoder · DUTSMAE: .157 -> .125Improved
use smoothness loss instead of cross-entropy loss l_c (ablation)Noise-aware encoder-decoder · DUTSS-measure: .813 -> .828Improved
use smoothness loss instead of cross-entropy loss l_c (ablation)Noise-aware encoder-decoder · DUTF-measure: .597 -> .701Improved
add smoothness loss to noise-aware model (ablation)Noise-aware encoder-decoder · training stabilitycollapseImproved
train with alternating back-propagation instead of cVAE (ablation)Noise-aware encoder-decoder · DUTSS-measure: .771 -> .828Improved
add noise generator when training on clean labels (ablation)f1* encoder-decoder · DUTSS-measure: .840 -> .861Improved
train on one noisy label per image (RBD) vs three labels (ablation)Noise-aware encoder-decoder · DUTSS-measure: .828 -> .824Mixed
use BASNet outputs as noisy labels (boosting) (ablation)BASNet · DUTSS-measure: .870Improved

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