Optimized hybrid attention enabled deep learning framework for yield monitoring and post-harvest management

India’s major economy relies on the contribution of agriculture particularly, vegetable crops yet challenges, such as a rapidly growing population, limited farmland, and frequent plant diseases threaten food security. In this regard, this research proposes an Optimized Attention-enabled EfficientNet B7 Convolutional Network to monitor the agricultural yields, like vegetable crops and to achieve sustainable agriculture. The architecture employs the EfficientB7 model along with the strengths of convolutional layers to effectively capture spatial dependencies between agronomic attributes, which enables the system to accurately classify crop diseases and consequently suggest pesticides for increasing production. The integration of Hybrid attention and Bright Attractive Social Interaction optimization further improves classification performance by focusing on relevant features and optimizing learning parameters effectively. The detection model showcases higher generalizability in disease detection and the performance is showcased for different datasets including the vegetable crops. The proposed model under comparison evaluation demonstrated accuracy of 97.41%, recall of 97.98%, and precision of 95.42% on the 80% training percentage, positioning it as an effective choice for increasing agricultural production. For the vegetable yield prediction, the accuracy of 97.4%, precision of 95.42%, and recall of 97.98%, respectively.