Automated detection of ocular diseases through retinal fundus images has become increasingly critical in the field of ophthalmology, as it aids in early diagnosis and management of vision-threatening conditions. This study presents a Convolutional Neural Network (CNN)-based model designed for multi-label classification of ocular diseases using the Ocular Disease Intelligent Recognition (ODIR-5K) dataset. The proposed model integrates depthwise separable convolutions and a modified inception block to enhance feature extraction while maintaining computational efficiency. Extensive data preprocessing, including normalization and data augmentation, was performed to improve model generalization. The model attained an accuracy of 98.94%, surpassing current baseline models in the domain. This approach addresses critical challenges such as class imbalance and coexisting disease conditions within a single sample, making it suitable for real-world clinical applications. Our findings suggest that this model has significant potential for deployment in automated ocular disease screening, particularly in resource-limited settings where access to specialized ophthalmic diagnostics is constrained.