The Do's and Don'ts for CNN-based Face Verification
This work addresses practical challenges in face verification for researchers and practitioners, but appears incremental as it explores existing questions without introducing a new method.
The paper tackles critical questions in face recognition research, such as training on still images for video performance and the impact of dataset depth versus width, by training CNNs on datasets like CASIA-WebFace and testing on YouTube-Faces, but does not report specific performance numbers or results.
While the research community appears to have developed a consensus on the methods of acquiring annotated data, design and training of CNNs, many questions still remain to be answered. In this paper, we explore the following questions that are critical to face recognition research: (i) Can we train on still images and expect the systems to work on videos? (ii) Are deeper datasets better than wider datasets? (iii) Does adding label noise lead to improvement in performance of deep networks? (iv) Is alignment needed for face recognition? We address these questions by training CNNs using CASIA-WebFace, UMDFaces, and a new video dataset and testing on YouTube- Faces, IJB-A and a disjoint portion of UMDFaces datasets. Our new data set, which will be made publicly available, has 22,075 videos and 3,735,476 human annotated frames extracted from them.