C-YOLOX: Improved YOLOX Model for Tomato Diseased Leaves Recognition

Different tomato growing stages are more likely to experience tomato diseases, and manual diagnosis requires a lot of work. To address the artificial recognition method's high rate of mistakes and poor accuracy, We proposed an improved C-YOLOX model built on the YOLOX-S model and applied it to healthy tomato leaves and 9 different tomato species Leaf diseases for training and identification. Firstly, we close the data augmentation in the last ⌊total epochs * 0. 25⌋ epochs to avoid the over-fitting problem. Secondly, the YOLOX-S model is further enhanced by adding a Coordinate attention mechanism module to the backbone network to better extract image feature information. Finally, we replace the silu activation function with the hard swish activation function to reduce computational costs. The experiments show that the revised C-YOLOX algorithm's best mAP50_95 value was 84.42%, 4.09 percentage points higher than it was before the change, demonstrating its efficacy. The upgraded C-YOLOX algorithm performs better.