22022.5.9 frames/segmentation/ 1. Images/image_0,1,2,3,... png 2. Masks/image_0,1,2,3,...png 导入adam改变了:如下： from keras.optimizers import adam_v2 adam = adam_v2.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08) Conv2DTranspose:逆卷积，由特征图求原来的输入。 concatenate：融合 文件夹浏览的 os.walk() https://blog.csdn.net/qq_37344125/article/details/107972463?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522165210122716782391853857%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=165210122716782391853857&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~top_positive~default-1-107972463-null-null.142^v9^control,157^v4^control&utm_term=os.walk&spm=1018.2226.3001.4187