Optimized YOLOVX Detection Method for Tobacco Debirs

Tobacco filaments are the main source of fuel for cigarettes and playa role in providing smoke and flavor. However, when tobacco contains debris, it can pose some health risks. Common tobacco debris includes tobacco paste, paper shavings, and mouthpiece residue. These debris can be inhaled into the human body during smoking, causing respiratory discomfort or other health problems. Therefore, accurate debris detection is an important issue that needs to be addressed in the tobacco processing process. In order to improve the detection accuracy of debris, we propose an optimized YOLOVX tobacco debris detection algorithm in this paper. First, we take a picture of the cigarette debris image to capture and label it. Then, in the backbone part, the SPP module is replaced by the ASPP module to suppress the irrelevant information in the image, which in turn improves the ability of the network to learn the region of interest. In the feature fusion part of the PAN et, the AFF module is used to make the fused feature map have both deep semantic information and shallow positional information to improve the detection accuracy of the tobacco clutter. The experiments on the tobacco clutter dataset show an improvement of 3.08% compared to the original YOLOX, which provides more accurate detection results for tobacco clutter.