Generally, all layers in Keras need to know the shape of their inputs in ... The answer was: (40, 40, 32), so we can keep downsampling.
確定! 回上一頁