Keypoints detection of Industrial manipulation gestures is a crucial need in a human-machine collaboration system. A high number of hand gesture keypoint detection problems are being presently solved. However, the accuracy as well as the reliability of hand gesture keypoints detection approaches still face challenges when they need to deal with more detailed hand manipulation behaviors such as gestures in the industrial manipulation. To solve the issue, a deep learning-based keypoint detection method for industrial manipulation gestures is proposed in this paper. Moreover, an industrial manipulation gesture dataset is established and a visualization software is designed for displaying the detection results. In addition, hand gesture keypoints detection experiments on the InterHand 2.6 M public dataset and the proposed industrial manipulation gesture dataset are conducted respectively. The experimental results show that the method can detect the keypoint position with high accuracy and can effectively detect the keypoints of industrial manipulation gestures.