tdf1995
tdf1995
![image](https://user-images.githubusercontent.com/39071957/67253449-42921b80-f4aa-11e9-8b34-5a02e5a52b2d.png) 我是这样放置的,train里放图片格式png,trainannot放标注文件格式png,还是不对,原本应该是4类 Epoch 1/30 Found 5484 images belonging to 2 classes. Found 2742 images belonging to 1 classes. Found 2742 images belonging to 1 classes. Found 5484 images belonging to...
我已经将num_classes设置为我需要的4,你的意思是在 dp = data_preprocess(train_path=train_path,image_folder=image_folder,label_folder=label_folder, valid_path=valid_path,valid_image_folder=valid_image_folder,valid_label_folder=valid_label_folder, img_rows=img_rows, img_cols=img_cols, flag_multi_class=flag_multi_class, num_classes=4) 将num_classes设置为2吗?我试了之后出现了错误: Error when checking target: expected conv2d_24 to have shape (512, 512, 4) but got array with shape (512, 512, 2)
按理说应该不是这些问题,因为我跑CamVid的数据集也是这个问题,你在跑CamVid的数据集时会打印12个类的图像信息吗?
能不能给我看下你训练CamVid的参数设置和log信息
我拼接成533*534了,还是不行,还有会出现 assert width > height AssertionError 请问是什么原因