Summary of Masked Face Recognition with Generative-to-discriminative Representations, by Shiming Ge et al.
Masked Face Recognition with Generative-to-Discriminative Representations
by Shiming Ge, Weijia Guo, Chenyu Li, Junzheng Zhang, Yong Li, Dan Zeng
First submitted to arxiv on: 27 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 The proposed unified deep network aims to improve masked face recognition by learning generative-to-discriminative representations. The network is split into three modules, with each module pre-trained on synthetic masked faces in a greedy manner. The first module uses a generative encoder pretrained for face inpainting and finetunes it to represent masked faces as category-aware descriptors. This helps provide occlusion-robust representations that can mitigate the effects of diverse masks. The second module is a multi-layer convolutional network that converts these category-aware descriptors into identity-aware vectors, using relation knowledge from an off-the-shelf face recognition model for supervision. The final module consists of a fully-connected layer and softmax layer that serves as a feature classifier, fine-tuned to identify the reformed identity-aware vectors. Experiments on synthetic and realistic datasets demonstrate the effectiveness of this approach in recognizing masked faces. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a way to recognize masked faces by learning generative-to-discriminative representations. It uses a special kind of computer model that is good at fixing broken pictures, and then tweaks it to be good at recognizing people’s faces even when they’re hidden behind masks. This helps make sure the recognition is accurate even if the mask is different or covers up parts of the face. The researchers tested their approach on fake and real photos and showed that it works well. |
Keywords
» Artificial intelligence » Convolutional network » Encoder » Face recognition » Mask » Softmax
