Anomaly identification of critical power plant facilities based on YOLOX-CBAM

Aiming at the abnormal phenomena of steam leakage, oil leakage and water leakage in the key facilities of the power plant, this paper proposes an anomaly detection method based on YOLOX and improves the network and parameters, and the improved network can be better applied to the anomaly detection of power plants. YOLOX is an improved version of the YOLO series, using Anchor-Free, Mosaic, SimOTA and other methods to improve the network. This article adds CBAM (Convolutional Block Attention Module) on this basis and the final experiment shows that the mAP of the YOLOX-CBAM model is 2.7% higher than that of YOLOX and the detection speed is only 1.4 FPS slower.