AI Paper Atlas

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

REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets

Angelina Wang, Arvind Narayanan, Olga Russakovsky

ECCV 2020 · Vision-language and multimodal › Human and embodied modeling · approach: Scientific domain vision datasets · 73 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
male instead of female person in COCO image areaCOCO person annotations · COCO person prominencefraction of image area: .138 -> .268Improved
male instead of female person, distance from centerCOCO person annotations · COCO person prominencedistance from image center: .510 -> .363Improved
darker skin tone of personCOCO person annotations · COCO skin tone trendperson size: p = 2.11e-7Hurt
darker skin tone of personCOCO person annotations · COCO skin tone trenddistance from center: p = .014Hurt
same-class vs different-class same-supercategory object pairs in feature spaceResNet-110 CIFAR-10 features · COCO appearance feature distancemean feature distance: 5.91 -> 6.24Improved
unrelated object pairs vs same-supercategory pairs in feature spaceResNet-110 CIFAR-10 features · COCO appearance feature distancemean feature distance: 6.24 -> 6.48Improved
add word-embedding supercategory binningOpenImages labels · OpenImages to COCO supercategoriesappropriate mappings of 50: 44 of 50Improved
darker skin tone of person, regional origin of imagesCOCO images of people · COCO geography by peopleshare from North America and Europe (%): 85.5 -> 58.2Hurt
darker skin tone of person, urban locationCOCO US images · COCO US urban shareshare in urban area (%): 84.4 -> 92.7Improved
New York City instead of Northern CaliforniaBDD100K · BDD100K weathersnowy image share (%): 0.3 -> 10Improved
unidentifiable gender person in COCO labeled male by annotatorsCOCO gender labels · COCO unidentifiable gendershare labeled male (%): 77Hurt
unidentifiable gender person in OpenImages labeled male by annotatorsOpenImages gender labels · OpenImages unidentifiable gendershare labeled male (%): 69Hurt

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