Fair Attribute Classification through Latent Space De-biasing
By Vikram V. Ramaswamy, Sunnie S. Y. Kim, Olga Russakovsky. 2020
193 citations recorded by the source corpus.
Fairness in visual recognition is becoming a prominent and critical topic of discussion as recognition systems are deployed at scale in the real world. Models trained from data in which target labels are correlated with protected attributes (e.g., gender, race) are known to learn and exploit those correlations. In this work, we introduce a method for training accurate target classifiers while mitigating biases that stem from these correlations. We use GANs to generate realistic-looking images, and perturb these images in the underlying latent space to generate training data that is balanced for each protected attribute. We augment the original dataset with this generated data, and empirically demonstrate that target classifiers trained on the augmented dataset exhibit a number of both quantitative and qualitative benefits. We conduct a thorough evaluation across multiple target labels and protected attributes in the CelebA dataset, and provide an in-depth analysis and comparison to existing literature in the space. Code can be found at https://github.com/princetonvisualai/gan-debiasing.
Subfields: Computer Vision and Pattern Recognition, AI Ethics, Fairness & Society
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Daniel Cohen-Or
- Rinon Gal
- Yotam Nitzan
- Ofir Brenner
- Luc Van Gool
- Angelina Wang
- Eli Shechtman
- Olga Russakovsky
In AI Ethics, Fairness & Society
- Angelina Wang
- Dora Zhao
- Vikram V. Ramaswamy
- Olga Russakovsky
- Yuta Nakashima
- Yusuke Hirota
- Noa García
Citations in the corpus
- Understanding and Evaluating Racial Biases in Image Captioning (2021, background): “Proposed solutions to dataset bias include new data collection approaches [38, 29], manual data cleanup [75, 77], synthetic data generation [59, 19, 61] – or, in extreme cases, even withdrawing the dataset after insurmountable biases have been identified [58].”
- LARGE: Latent-Based Regression through GAN Semantics (2021, background): “Others enriched a real dataset to improve performance in the low-data domain [4,14,66,80] or on biased or unbalanced data [16,20,32,43,50].”
- Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation (2022, background): “Recent work has proposed leveraging Generative Adversarial Networks (GANs) to create synthetic examples to help train models, especially for facial datasets [102, 105, 111], 11 but some of these visual results can feel akin to the harmful performance of blackface.”