A Robust Framework for Real-Time Iris Landmarks Detection Using Deep Learning

Iris detection and tracking plays a vital role in human–computer interaction has become an emerging field for researchers the last two decades. Typical applications such as virtual reality, augmented gaze customer behavior, controlling computers, handheld embedded devices need accurate precise of iris landmarks. A significant improvement been made so far tracking. However, landmarks real-time with high accuracy is still challenge computationally expensive task. This also accompanied lack publicly available dataset annotated article presents benchmark robust framework localization key landmark points to extract better accuracy. number training sessions have conducted MobileNetV2, ResNet50, VGG16, VGG19 over dataset, ImageNet weights are used model initialization. The Mean Absolute Error (MAE), loss, size measured evaluate validate proposed model. Results analyses show that outperforms other methods on selected parameters. MAEs 0.60, 0.33, 0.35, 0.34; average decrease 60%, reduction response time 75% compared models. We collected images eyes them help algorithm. generated research purposes. contribution this more diminutive prediction landmarks, along provided annotations.