Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Facial Action Units From Web Images With Scalable Weakly Supervised Clustering
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 |
|---|---|---|---|
| Re-annotate weak labels with WSC instead of using weak labels directly | AlexNet trained on EmotioNet · EmotioNet test | F1: 53.6 -> 56.1 | Helped |
| Re-annotate weak labels with WSC instead of using weak labels directly | AlexNet annotations · EmotioNet annotation agreement | agreement (%): 89.2 -> 91.3 | Helped |
| Add WSC re-annotated images to training | AlexNet trained with 25k human-labeled images · EmotioNet test | F1: 56.9 -> 58.1 | Helped |
| Add WSC re-annotated images to training | DRML trained with 25k human-labeled images · EmotioNet test | F1: 57.0 -> 58.1 | Helped |
| Add WSC re-annotated images to training | AlexNet trained with 25k human-labeled images · EmotioNet test | S score: 75.5 -> 76.5 | Helped |
| Add WSC re-annotated images to training | DRML trained with 25k human-labeled images · EmotioNet test | S score: 75.7 -> 75.9 | Helped |
| Use WSC re-annotated rather than weakly annotated training images | AlexNet trained on 20k images · EmotioNet test | F1: 47.5 -> 49.7 | Helped |
| Use WSC re-annotated rather than weakly annotated training images | AlexNet trained on 200k images · EmotioNet test | F1: 48.5 -> 50.8 | Helped |
| Use WSC re-annotated rather than weakly annotated training images | AlexNet trained on 400k images · EmotioNet test | F1: 47.9 -> 52.1 | Helped |
| Use WSC re-annotated rather than weakly annotated training images | AlexNet trained on 1M images · EmotioNet test | F1: 49.3 -> 55.5 | Helped |
| Increase weakly annotated training set from 200k to 400k images | AlexNet trained with weak labels · EmotioNet test | F1: 48.5 -> 47.9 | Hurt |
| Increase WSC re-annotated training set from 20k to 1M images | AlexNet trained with WSC labels · EmotioNet test | F1: 49.7 -> 55.5 | Helped |
Closest papers by meaning
- EmotioNet: An Accurate, Real-Time Algorithm for the Automatic Annotation of a Million Facial Expressions in the Wild (CVPR 2016 · similarity 0.89)
- Weakly Supervised Facial Action Unit Recognition Through Adversarial Training (CVPR 2018 · similarity 0.89)
- Weakly-Supervised Deep Convolutional Neural Network Learning for Facial Action Unit Intensity Estimation (CVPR 2018 · similarity 0.87)
- Exploiting Semantic Embedding and Visual Feature for Facial Action Unit Detection (CVPR 2021 · similarity 0.87)
- Classifier Learning With Prior Probabilities for Facial Action Unit Recognition (CVPR 2018 · similarity 0.87)
- Distribution Matching for Multi-Task Learning of Classification Tasks: A Large-Scale Study on Faces & Beyond (AAAI 2024 · similarity 0.86)
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.