2017/12/07 by Alex Levinshtein, Levinshtein, Alex, Edmund Phung +3
Computer Science · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Gaze Tracking and Assistive Technology #Ocular Diseases and Behçet’s Syndrome #Retinal and Optic Conditions
paper · pdf · doi:10.48550/arxiv.1712.02822
openalex publication_date 2017/12/07 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
We propose a new cascaded regressor for eye center detection. Previous\nmethods start from a face or an eye detector and use either advanced features\nor powerful regressors for eye center localization, but not both. Instead, we\ndetect the eyes more accurately using an existing facial feature alignment\nmethod. We improve the robustness of localization by using both advanced\nfeatures and powerful regression machinery. Unlike most other methods that do\nnot refine the regression results, we make the localization more accurate by\nadding a robust circle fitting post-processing step. Finally, using a simple\nhand-crafted method for eye center localization, we show how to train the\ncascaded regressor without the need for manually annotated training data. We\nevaluate our new approach and show that it achieves state-of-the-art\nperformance on the BioID, GI4E, and the TalkingFace datasets. At an average\nnormalized error of e < 0.05, the regressor trained on manually annotated data\nyields an accuracy of 95.07% (BioID), 99.27% (GI4E), and 95.68% (TalkingFace).\nThe automatically trained regressor is nearly as good, yielding an accuracy of\n93.9% (BioID), 99.27% (GI4E), and 95.46% (TalkingFace).\n