Key takeaways
- Meta has launched FACET, a new benchmark designed to assess the fairness of computer vision models across various tasks including classification and detection.
- The FACET dataset comprises 32,000 images featuring 50,000 individuals, with detailed demographic annotations provided by expert human annotators.
- Meta's DINOv2 model, trained through self-supervised learning, is now available under the Apache 2.0 licence, along with a set of dense prediction models for enhanced flexibility in applications.
- The introduction of FACET aims to address challenges in evaluating fairness in computer vision, particularly issues related to mislabeling and demographic biases.
- Evaluations of DINOv2 using the FACET benchmark have highlighted specific performance nuances, indicating the model's varied effectiveness in different demographic contexts.
Meta on Thursday announced that DINOv2, its computer vision model trained through self-supervised learning to produce universal features, is now available under the Apache 2.0 license.
The company further added that it is also releasing a collection of DINOv2-based dense prediction models for semantic image segmentation and monocular depth estimation, giving developers and researchers even greater flexibility to explore its capabilities on downstream tasks.
Alongside DINOv2's announcement, Meta introduced FACET (FAirness in Computer Vision EvaluaTion), a new comprehensive benchmark for evaluating the fairness of computer vision models across classification, detection, instance segmentation, and visual grounding tasks.
Meta said this move is in response to the challenging nature of benchmarking fairness in computer vision, which has often been hampered by potential mislabeling and demographic biases.
Meta in their blog post said that the FACET’s dataset is made up of 32,000 images containing 50,000 people, which is labeled by expert human annotators for demographic attributes. Additionally, FACET also contains person, hair, and clothing labels for 69,000 masks from SA-1B.
Meta evaluated Dinov2 using FACET which revealed nuances in its performance, particularly in gender-biased classes.
Meta said it hopes that FACET can become a standard fairness evaluation benchmark for computer vision models and help researchers evaluate fairness and robustness across a more inclusive set of demographic attributes. For the same purpose, Meta released the FACET dataset and a dataset explorer.
To make FACET effective, Meta said it hired expert reviewers to manually annotate person-related demographic attributes like perceived gender presentation and perceived age group as well as correlating visual features like perceived skin tone, hair type, and accessories.
Additionally, the dataset includes labels for person-related classes like "basketball player" and "doctor," as well as attributes related to clothing and accessories.
“By releasing FACET, our goal is to enable researchers and practitioners to perform similar benchmarking to better understand the disparities present in their own models and monitor the impact of mitigations put in place to address fairness concerns,” Meta wrote in the blog post.
Spotted an error? Tell our editors. Read our editorial standards and corrections.






