keras+opencv+openpose实现基于CNN的人体动作预测。 训练部分用到了colab，使用的数据集是unity的视频数据集 数据集链接：http://www.aislab.org/index.php/en/mvar-datasets/ data_op文件对数据进行openpose处理 image_data_prepare文件制作data和label的npy文件 action_detect文件为模型训练及可视化 model_video_test为测试可视化
边学习边做的小项目，不准，很菜。
Keras+opencv+openpose realizes human motion prediction based on CNN. Colab was used in the training section. The dataset used is the video dataset of unity3D Dataset links: http://www.aislab.org/index.php/en/mvar-datasets/. The data_op file processes the data with openpose. Image_data_prepare files make npy files for data and label. The action_detect file is model training and visualization. Model_video_test is visual for testing.