Abstract: Targeting the problems of poor real-time performance and weak small target detection ability existing in current methods for bird’s nest detection on transmission lines using UAV imagery,the paper proposes the nest detection method for lightweight YOLOv3 transmission line based on deep separable convolution. Firstly,Mosaic data augmentation method is used to enhance the dataset and improve the number of small targets in training set. Then deep separable convolution is used to replace part of standard convolution in main feature extraction network to improve the speed of a detection network and reduce the number of network parameters,reducing the weight file size. In addition,PANet is used instead of FPN to further improve the ability of feature fusion and enhance the detection ability of the small targets. Finally,label smoothing is used for training to solve the problems of the overconfidence and overfitting of the network caused by a small number of labeling errors. The UAV inspection video of a power supply bureau is cut into an image to make a data set. The proposed algorithm is compared with the original YOLOv3 algorithm,and the ablation experiment is carried out. The results show that the proposed algorithm improves gradually the speed and accuracy of the model.