Model is not working for different dataset
sanahtech opened this issue · comments
SRGAN model is working fine for given DIV2K_train_hr dataset. But I tried to train the same model for a different dataset. It does not work.
following is the error.
File "init.py", line 83, in
train(FLAGS)
File "/home/student/usama_lahore/Memona/perceptual-reflection-removal-master/SRGANmodelnewdata/SRGAN-Keras-Implementation-master/train.py", line 21, in train
srgan_model.train(epochs, save_interval = save_interval, batch_size = batch_size)
File "/home/student/usama_lahore/Memona/perceptual-reflection-removal-master/SRGANmodelnewdata/SRGAN-Keras-Implementation-master/models/SRGAN.py", line 225, in train
real_dis_loss = self.discriminator.train_on_batch(hr_imgs, real)
File "/usr/local/lib/python3.5/dist-packages/keras/engine/training.py", line 1211, in train_on_batch
class_weight=class_weight)
File "/usr/local/lib/python3.5/dist-packages/keras/engine/training.py", line 751, in _standardize_user_data
exception_prefix='input')
File "/usr/local/lib/python3.5/dist-packages/keras/engine/training_utils.py", line 128, in standardize_input_data
'with shape ' + str(data_shape))
ValueError: Error when checking input: expected input_3 to have 4 dimensions, but got array with shape (16, 1)
I have trained new GANs using this repo with a folder of images. How is the shape of your data (16,1)? Could you give more info on the experiment you are doing?
norm_hr = self.high_reso_imgs[indx_high] / 127.5 - 1
print(norm_hr.shape)
in gen_pipeline(self, batch_size=16) function the above line prints (16, ).
while it should print (16,256,256,3)