This paper presents a method for detecting cataracts and glaucoma using transfer learning with MobileNetV1 and MobileNetV2. The authors aim to address the challenges of early and accurate detection of these eye conditions, which can lead to blindness. The proposed model is designed to classify images into three categories: cataract, glaucoma, and normal. The model architecture includes depth-wise separable convolutions and global pooling layers, enhancing its ability to learn and generalize from the data. The experiments are conducted using publicly available datasets, and the results show that the proposed model achieves higher accuracy compared to other models. The study also evaluates the model's performance on a benchmark dataset and compares it with previous works, demonstrating its effectiveness and efficiency in detecting eye diseases. The key contributions of the paper include the customized architecture for eye disease detection and the robust multi-label classification approach, which enhances diagnostic capabilities. The proposed model is particularly useful for resource-constrained environments and mobile devices due to its lightweight nature.This paper presents a method for detecting cataracts and glaucoma using transfer learning with MobileNetV1 and MobileNetV2. The authors aim to address the challenges of early and accurate detection of these eye conditions, which can lead to blindness. The proposed model is designed to classify images into three categories: cataract, glaucoma, and normal. The model architecture includes depth-wise separable convolutions and global pooling layers, enhancing its ability to learn and generalize from the data. The experiments are conducted using publicly available datasets, and the results show that the proposed model achieves higher accuracy compared to other models. The study also evaluates the model's performance on a benchmark dataset and compares it with previous works, demonstrating its effectiveness and efficiency in detecting eye diseases. The key contributions of the paper include the customized architecture for eye disease detection and the robust multi-label classification approach, which enhances diagnostic capabilities. The proposed model is particularly useful for resource-constrained environments and mobile devices due to its lightweight nature.