Automated Crop Inspection

Conventional methods in the field of agriculture are often time-consuming and inaccurate, because of their high dependency on human intervention. Non-smart crop inspection techniques are done by the visual inspection of the farmer to crops, and by using chemical analysis in some cases of fruit or leaf disease. This process not only demands a substantial amount of time, but also is inefficient and unreliable. The current research focuses on providing AI-based solutions to automate the two tasks of fruit ripeness detection and leaf disease classification. For the first module, the object detection model, fine-tuning with Yolov8m and Faster-RCNN was performed, and the results were compared. As for the second module, the multi-class image classifier, transfer learning with ResNet50 and VGG16 was conducted, and results were documented and analyzed. Experiments have been carried out using different datasets to assess the performance and generalizability.