Research map / Label-efficient and robust learning / Noisy labels
Deep Unsupervised Saliency Detection: A Multiple Noisy Labeling Perspective
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · THUR | Fβ: .7115 -> .7322 | Helped |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · PASCAL-S | Fβ: .8391 -> .8422 | Helped |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · MSRA-B | Fβ: .8973 -> .8770 | Hurt |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · ECSSD | Fβ: .8908 -> .8783 | Hurt |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · DUT | Fβ: .7360 -> .7156 | Hurt |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · SED2 | Fβ: .8745 -> .8380 | Hurt |
| Train with explicit noise modeling and joint latent saliency prediction from multiple noisy labels rather than clean human labels | Deep unsupervised saliency detector · SOD | Fβ: .8235 -> .7976 | Hurt |
| Use explicit noise modeling and iterative updates instead of directly training on noisy saliency labels | Deep unsupervised saliency detector · Seven benchmark datasets | Fβ | Helped |
| Average the four unsupervised saliency maps into one pseudo-label rather than using each map as a separate noisy label | BL2 · Seven benchmark datasets | Fβ | Hurt |
| Add successive noise-model updates after initializing noise variance to zero (ablation) | Deep unsupervised saliency detector · Seven benchmark datasets | MAE | Helped |
| Make the noise distribution depend on the labeller (ablation) | Deep unsupervised saliency detector · Saliency detection | performance | No clear effect |
Closest papers by meaning
- DeepUSPS: Deep Robust Unsupervised Saliency Prediction via Self-supervision (NeurIPS 2019 · similarity 0.93)
- Promoting Saliency From Depth: Deep Unsupervised RGB-D Saliency Detection (ICLR 2022 · similarity 0.91)
- Deep Saliency With Encoded Low Level Distance Map and High Level Features (CVPR 2016 · similarity 0.91)
- Weakly Supervised Salient Object Detection Using Image Labels (AAAI 2018 · similarity 0.90)
- Unsupervised Domain Adaptive Salient Object Detection through Uncertainty-Aware Pseudo-Label Learning (AAAI 2022 · similarity 0.90)
- Unsupervised Salient Instance Detection (CVPR 2024 · similarity 0.89)
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.