Abstract: To solve the problems of high missed detection rate, frequent misjudgment and difficult model fitting in infrared detection of gas leakage, a YOLOv8-CBAM method with increasing the CBAM attention mechanism module and optimizing the loss function to Focaler-loU was proposed to improve the accuracy and stability of infrared detection of gas leakage. Experimental results show that the detection accuracy of the improved YOLOv8-CBAM convolutional neural network is significantly better than that of other mainstream object detection network models, and the average detection accuracy of the test set is as high as 98.82%. The average number of detection frames per second is 109.9frames, which meets the real-time and accuracy requirements of actual gas leakage detection tasks.