Cucumber farming plays a crucial role in Bangladesh's agricultural economy, significantly contributing to vegetable production. However, diseases like Downy Mildew, Bacterial Wilt, Anthracnose, and Belly Rot threaten crop quality and lead to financial losses. Effective management of these diseases is crucial for maintaining high crop yields and ensuring economic stability. Current disease identification methods are labor-intensive, costly, and dependent on expert knowledge. To address these challenges, this study presents an automated disease categorization system using advanced transfer learning models, including MobileNet, ShuffleNet, InceptionResNet-v2, AlexNet, and Ensemble Transfer Learning (ETL). Utilizing a dataset of 8,000 well-annotated images, enhanced through preprocessing techniques like GLRM feature extraction and InfoMAX-GAN feature selection, the ETL model achieved a remarkable accuracy of 99.79% and an F1-score of 99.78%. We also developed a user-friendly web tool for real-time diagnostics, providing invaluable support to farmers and agricultural specialists. This tool enables rapid disease detection, leading to increased productivity and a transformative approach to cucumber disease management. By reducing the reliance on chemical interventions, it promotes sustainable farming practices and safeguards human and animal health.