SqueezeNet-Based Model for Subject Identification from Off-Angle Iris Image

Identification of subjects using biometric features such as the iris has gained popularity over the years as a reliable and secure method for authentication. However, most iris recognition systems require high-quality images captured at a specific angle, limiting their usability in real-world scenarios where images may not be of good quality or captured at the ideal angle. It is still possible to take off-angle iris photos even with qualified operators and cooperative participants. To extract features for iris identification, SqueezeNet a CNN-based architecture has been proposed. Trials are performed utilizing image data taken from different sections of the human eye to determine which part of the eye is best suited for the recognition system using CNN. Several gaze angles are examined individually in trials to determine their effects on recognition skills and picture data based on different gaze angles.