Loading datasets...
Release Date: 10/2/2026
Format: PNG, JSON, CSV, TXT
Size: 3.88 GB
Contains metadata files associated with the four primary FHIBE variants: full resolution, downsampled, face crop only, and aligned face crop. You can find FHIBE's data card and transparency documentation here: fairnessbenchmark.ai.sony/data-card The metadata includes: - Demographic information of annotators and QA reviewers. - Summary CSV files containing annotations extracted from the corresponding JSON files, enabling annotation analysis without loading individual JSON files. - Aggregated pre-computed evaluation scores and analysis results for the downsampled and aligned face crop variants. - Pre-computed model predictions and evaluation protocols for face verification, including genuine and imposter pairs, for the aligned face crop variant. - Additional pre-computed model prediction and evaluation results for face parsing and face verification for the aligned face crop variant. - Extracted face masks (`.png`) following the CelebAMask-HQ format for the face crop variants. Summary CSV files are located in the corresponding dataset directories: - data/processed/fhibe_downsampled/fhibe_downsampled.csv - data/processed/fhibe_fullres/fhibe_fullres.csv - data/processed/fhibe_face_crop_align/fhibe_face_crop_align.csv - data/processed/fhibe_face_crop_only/fhibe_face_crop_only.csv Face masks are located under: data/processed/<dataset_variant>/masks/CelebAMask-HQ_format/ Note: Aggregated pre-computed evaluation scores and analysis results are not available for the full-resolution and face crop-only variants. For support related to this dataset, including consent revocation, please contact [email protected]
Licensing
FHIBE Term of Use
Restrictions/Special Constraints
As set out in Section 2.1 of the Terms of Use, FHIBE is primarily an evaluation dataset and may only be used as a training dataset for the development of bias diagnostic and mitigation tools or methods.
Forbidden Usage
As a user of FHIBE, you may not, and may not permit or assist others, to:
Attempt to re-identify any individuals in FHIBE.
Attempt to infer, predict, or label any sensitive or objectionable attributes to the Licensed Images, such as race or ethnicity, gender or sexual orientation, political opinions, religion or religious beliefs, genetic data, propensity towards crime, personality, attractiveness, etc., except in connection with detecting bias in model analysis.
Use biometric data in any Licensed Images to perform any Processing activities (as defined in the Data Sharing Addendum in Exhibit B of the Terms of Use) unrelated to bias evaluation and mitigation, including without limitation facial recognition.
Use FHIBE in any way that causes reputational harm to the individuals in the Licensed Images or other parties.
Use FHIBE in any manner that violates any applicable law, such as privacy or security laws or regulations as well as including accessing FHIBE from a country that is subject to a U.S. Government embargo.
Use FHIBE for any law enforcement, military, arms or surveillance purpose or any other purpose other than the Purpose.
Use FHIBE for evaluating AI systems that (are banned in, or violate restrictions of, the United States, the European Union and in your applicable jurisdiction.
For more restrictions please refer to the Terms of Use and Acceptable Use Policy.
Ethical Review
Data collection commenced after April 23, 2023, following Institutional Review Board (IRB) approval from WCG Clinical, Inc. (study number 1352290). All participants provided informed consent, and image subjects additionally consented to their identifiable images being included in the dataset.
Intended Use
FHIBE is intended to be used as a bias/fairness evaluation dataset for machine learning models. The diverse set of images and rich annotations enable evaluations over a wide range of computer vision tasks, including person detection and segmentation, face detection and verification, keypoint (pose) estimation, and visual question answering. Applications for which FHIBE has been used include surfacing gender biases in vision language model question answering and identifying hair style variability as a leading contributor to gender bias in popular face verification models.
fhibe_downsampled/ Contains: Aggregated pre-computed evaluation scores and analysis results. Demographic information of the annotators and QA reviewers involved in the annotation process. A summary CSV file containing all FHIBE downsampled annotations extracted from the corresponding JSON files. This file is located at:
data/processed/fhibe_downsampled/fhibe_downsampled.csv
It is particularly useful for performing analyses on the data annotations, without the need of loading and parsing the individual JSON annotation files.
Total files: 4 CSV files: 4
Number of rows in summary file fhibe_downsampled.csv: 11524 Number of columns in summary file fhibe_downsampled.csv: 78
fhibe_fullres/ Contains: Demographic information of the annotators and QA reviewers involved in the annotation process. A summary CSV file containing all FHIBE annotations extracted from the corresponding JSON files. This file is located at:
data/processed/fhibe_fullres/fhibe_fullres.csv
It is particularly useful for performing analyses on the data annotations, without the need of loading and parsing the individual JSON annotation files.
Note: No aggregated precomputed evaluation scores or analysis results are available for the full-resolution dataset.
Total files: 3 CSV files: 3
Number of rows in summary file fhibe_fullres.csv: 11524 Number of columns in summary file fhibe_fullres.csv: 78
fhibe_face_crop_align/ Contains: Aggregated pre-computed evaluation scores and analysis results. Demographic information of the annotators and QA reviewers involved in the annotation process. Pre-computed model predictions for face verification tasks. Face verification evaluation protocols, including genuine and imposter pairs. Additional pre-computed model prediction and evaluation result files for face parsing and face verification tasks A summary CSV file containing all annotations extracted from the corresponding JSON files.This file is located at:
data/processed/fhibe_face_crop_align/fhibe_face_crop_align.csv.
It is particularly useful for performing analyses on the data annotations, without the need of loading and parsing the individual JSON annotation files. Extracted face masks (.png) following the CelebAMask-HQ format. The masks are located at:
data/processed/fhibe_face_crop_align/masks/CelebAMask-HQ_format/
Total files: 119504 CSV files: 4 PNG files (masks): 119491 JSON files (model predictions and outputs): 8 TXT files (face verification protocol): 1
Number of rows in summary file fhibe_face_crop_align.csv: 8947 Number of columns in summary file fhibe_face_crop_align.csv: 75
fhibe_face_crop_only/ Contains: Demographic information of the annotators and QA reviewers involved in the annotation process. A summary CSV file containing all annotations extracted from the corresponding JSON files. This file is located at:
data/processed/fhibe_face_crop_only/fhibe_face_crop_only.csv.
It is particularly useful for performing analyses on the data annotations, without the need of loading and parsing the individual JSON annotation files. Extracted face masks (.png) following the CelebAMask-HQ format. The masks are located at:
data/processed/fhibe_face_crop_only/masks/CelebAMask-HQ_format/
Note: No aggregated pre-computed evaluation scores or analysis results are available for the face crop-only dataset.
Total files: 150856 CSV files: 3 PNG files (masks): 150853
Number of rows in summary file fhibe_face_crop_only.csv: 11524 Number of columns in summary file fhibe_face_crop_only.csv: 75
Compensated · similar languages
High-specularity optical edge-case dataset (425 assets) for benchmarking computer vision and autonomous navigation against severe specular reflections.
Versioned software-engineering corpus with code, configuration, workflows, and AI-tooling artifacts across frontend, backend, cloud, and infrastructure.
