In recent years, deep learning technology has achieved breakthrough development in the field of mechanical and electrical equipment fault diagnosis, but the research on the problem of motor vibration signal fault feature representation and model channel space correlation learning is still insufficient. Therefore, based on the existing research, this study proposes a motor fault diagnosis method based on multi-scale texture enhancement and dual attention mechanism. Firstly, by constructing multi-scale gray texture image (MGTI), the high fidelity conversion of vibration signal from one-dimensional time domain to two-dimensional texture space is realized. Secondly, by introducing the channel space dual attention module (CBAM), an improved residual network (CBRN) is designed to adaptively calibrate the feature response to focus on the key fault information. Finally, the experiments are conducted on the induction motor (IM) show that the accuracy of the eight types of motor fault diagnosis is 99.75%, which is up to 12.25% higher than the traditional fault diagnosis model (such as SVM). The experimental results confirm the feasibility and effectiveness of CBRN in the field of fault diagnosis.