AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Gender Artifacts in Visual Datasets

Nicole Meister, Dora Zhao, Angelina Wang, Vikram V. Ramaswamy, Ruth Fong, Olga Russakovsky

ICCV 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Human perception modeling · 38 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
reduce image resolution to 28x28 and grayscale (ablation)gender artifact model (ResNet-50) · COCOAUC (%): 61.9 (color) vs 51.9 (grayscale)Hurt
remove background (show person only) (ablation)gender artifact model · COCOAUC (%): 93.4 -> 92.7No clear effect
occlude person with rectangle, keep background (ablation)gender artifact model · COCOAUC (%): 93.4 -> 70.8Hurt
reduce person to bounding box only (no background) (ablation)gender artifact model · COCOAUC (%): 93.4 -> 58.0Hurt
use only mean RGB color features (ablation)logistic regression · COCO / OpenImagesAUC (%): 58.0 (COCO), 59.1 (OpenImages)No clear effect
use more random objects (10 -> 30) in object-presence classifier (ablation)logistic regression · COCOAUC (%): 58.0 -> 66.8Improved
train on white instead of black occlusion pixels (ablation)gender artifact model · COCO / OpenImagesAUC (%): largest change 1.5% COCO, 3.5% OpenImagesNo clear effect
balance the dataset by gender (ablation)gender artifact model · COCOAUC (%): 65.6 -> 63.6 (largest change)No clear effect
pretrain on PASS (no people) instead of ImageNet (ablation)gender artifact model · COCOAUC (%): 93.4 -> 89.9 (Full)No clear effect
adversarial gender debiasing (fairness through blindness)Wang et al. U-Net · COCO vallocation of removed pixels: 23.2% person vs 76.8% backgroundHurt

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