Research map / Vision-language and multimodal / Human and embodied modeling
Beyond Skin Tone: A Multidimensional Measure of Apparent Skin Color
William Thong, Przemyslaw Joniak, Alice Xiang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| measure skin hue in addition to skin tone on face datasets | CelebAMask-HQ and FFHQ · dataset skin color distribution | share of samples (%): light-red 46.95 vs dark-yellow 06.29 (CelebAMask-HQ); 52.82 vs 04.68 (FFHQ) | Improved |
| generative models trained on FFHQ | StyleGAN3 and P2 · generated skin color distribution | share light-red (%): FFHQ 52.82 -> StyleGAN3 53.05, P2 53.29 | Hurt |
| light vs dark skin tone faces in Twitter cropping | Twitter saliency cropping · CFD | preference (%): 60.73% preference for light skin | Hurt |
| red vs yellow skin hue faces in Twitter cropping | Twitter saliency cropping · CFD | preference (%): 58.25% preference for red | Hurt |
| light skin tone and red hue versus other groups in face verification | ArcFace, FaceNet, Dlib · LFW | accuracy: ArcFace light 95.52 vs dark 94.39 | Mixed |
| make skin tone lighter in male images (ablation) | AWS gender classification · CelebAMask-HQ male | accuracy: 94.82 -> 90.66 | Hurt |
| make skin hue redder or yellower (ablation) | gender classifiers · CelebAMask-HQ | gender accuracy: AWS All 97.88 -> 97.92 (+red) | No clear effect |
| make skin hue redder in non-smiling images (ablation) | Azure smile classification · CelebAMask-HQ non-smiling | accuracy: 80.84 -> 69.76 | Hurt |
| make skin tone darker in male images (ablation) | AWS gender classification · CelebAMask-HQ male | accuracy: 94.82 -> 97.73 | Improved |
Closest papers by meaning
- FACET: Fairness in Computer Vision Evaluation Benchmark (ICCV 2023 · similarity 0.87)
- Towards Racially Unbiased Skin Tone Estimation via Scene Disambiguation (ECCV 2022 · similarity 0.86)
- Consensus and Subjectivity of Skin Tone Annotation for ML Fairness (NeurIPS 2023 · similarity 0.86)
- Towards causal benchmarking of bias in face analysis algorithms (ECCV 2020 · similarity 0.84)
- Hyper-Skin: A Hyperspectral Dataset for Reconstructing Facial Skin-Spectra from RGB Images (NeurIPS 2023 · similarity 0.84)
- Leveraging Diffusion Perturbations for Measuring Fairness in Computer Vision (AAAI 2024 · similarity 0.82)
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.
