Summary of Facesaliencyaug: Mitigating Geographic, Gender and Stereotypical Biases Via Saliency-based Data Augmentation, by Teerath Kumar et al.
FaceSaliencyAug: Mitigating Geographic, Gender and Stereotypical Biases via Saliency-Based Data Augmentation
by Teerath Kumar, Alessandra Mileo, Malika Bendechache
First submitted to arxiv on: 17 Oct 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In this study, researchers introduce FaceSaliencyAug, an approach aimed at reducing gender bias in Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). The proposed method utilizes salient regions of faces detected by saliency maps to mitigate geographical and stereotypical biases in datasets. By randomly selecting masks from a predefined search space and applying them to the salient region of face images, FaceSaliencyAug enhances data diversity, leading to improved model performance and debiasing effects. The researchers quantify dataset diversity using Image Similarity Score (ISS) across five datasets, including Flickr Faces HQ (FFHQ), WIKI, IMDB, Labelled Faces in the Wild (LFW), UTK Faces, and Diverse Dataset. FaceSaliencyAug demonstrates superior diversity metrics as evaluated by ISS-intra and ISS-inter algorithms. The study also evaluates the effectiveness of the approach in reducing gender bias on CEO, Engineer, Nurse, and School Teacher datasets using Image-Image Association Score (IIAS). Results show a reduction in gender bias for both CNNs and ViTs, indicating the efficacy of FaceSaliencyAug in promoting fairness and inclusivity in computer vision models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The researchers created a new way to make computer vision models more fair by reducing biases. They called it FaceSaliencyAug. This method helps remove biases from datasets by using special maps that highlight important parts of faces. Then, it randomly adds masks to these areas and restores the original image. By doing this, the model becomes more diverse and accurate, which is good for fairness. The researchers tested their approach on many different datasets and showed that it works well. |
