Attention-Augmented SqueezeNet for Efficient Glaucoma Detection using Retinal Imaging

Glaucoma stands as a principal reason behind irreversible blindness which shows no warning signs at the beginning making diagnosis critical at an early stage. Retinal image manual assessment through traditional diagnosis methods demands extended time along with a significant likelihood of human-caused mistakes. The existing automated techniques suffer from problems during the feature extraction stage along with difficulties in precisely identifying essential areas. We have developed an attention-augmented SqueezeNet-based Convolutional Neural Network (CNN) for glaucoma detection to handle identified problems. The proposed model implements Fire Modules to boost its feature extraction abilities while using a reduced architectural design that minimizes operational expenses. Attention mechanisms enable better accuracy by selecting important glaucoma-affected areas within the image features. The added features produce a streamlined system which enhances both computation speed and decreases hardware necessity to support real-time clinical deployments. This model demonstrates high diagnostic capability for differentiating normal and glaucoma-affected retinal images thus establishing itself as a suitable early detection instrument for medical practitioners.