A Fruit and Vegetable Recognition Method based on AugMSR-CNN

The traditional manual settlement method is characterized by low efficiency, high labor and resource consumption, and significant negative impact on consumers' shopping experience. To effectively address the aforementioned issues, a fruit and vegetable recognition model based on the MSR-CNN algorithm is proposed in this paper. Additionally, an AugFPN feature pyramid is innovatively incorporated into the MSR-CNN algorithm to enhance the model's generalization ability. Experimental results demonstrate that the model, based on the algorithm proposed in this paper, achieves an average increase of 1.4 percentage points in AP value, greatly improving the recognition rate of fruits and vegetables.