An integrated approach for eye centre localization using deep networks and rectangular-intensity-gradient technique

Abstract

Eye centre localization plays a crucial role in computer vision applications like face recognition, gaze estimation, driver fatigue detection, liveness detection, etc. However, it is difficult to localize the eye centre due to the variations in pose, occlusion, illumination, specular reflection, rotation, scale, etc. This work proposes an integrated approach robust to the variations mentioned above and localizes the eye centre precisely. This integrated approach consists of (i) Faster RCNN deep learning model to detect the face and eyes; (ii) AlexNet is utilized for eye openness detection using transfer learning; (iii) Rectangular-intensity-gradient (RIG) is proposed for eye centre localization. Experimentations were performed on benchmark databases like GI4E and BioID. An accuracy of 97.50% and 94.25% for Nerr ≤ 0.05, 98.75% and 98.40% for Nerr ≤ 0.10, 99.64% and 99.45% for Nerr ≤ 0.25 was achieved across GI4E and BioID databases respectively. In addition, the robustness of the proposed approach was tested on AR and CAS-PEAL databases. The proposed approach of eye centre localization performs relatively better than existing state-of-the-art methods in terms of accuracy and computational time.