Eye Diseases Classification with Streamlit (Custom CNN, EfficientNetB7)

• Developed a deep learning system to classify retinal fundus images into 4 disease categories: Cataract, Diabetic

Retinopathy, Glaucoma, and Normal.

• Built and compared a scratch-designed CNN (89% accuracy) and a fine-tuned EfficientNetB7 (95% accuracy)

using TensorFlow/Keras.

• Deployed as an interactive Streamlit app for real-time predictions to assist early ophthalmic diagnosis.

• Tools Used: Python, TensorFlow/Keras, Streamlit, Transfer Learning (EfficientNetB7)