Error updating Conditional GAN to work on RGB images

General Discussion
#1

I’ve working with the example for Conditional GAN

It works great on grayscale images, however when I use dataset of RGB images it throws an error.
I’ve modified the variables related to the num_categories (6) and num_channels (3) to match my dataset, however there’s a problem with mismatched tensor sizes later on between the fake_image_and_labels and real_image_and_labels when concatenating.

Thanks in advance!

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_1421/662719055.py in <module>
      8 )
      9 
---> 10 cond_gan.fit(dataset, epochs=20)

/usr/local/lib/python3.8/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs)
     65     except Exception as e:  # pylint: disable=broad-except
     66       filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67       raise e.with_traceback(filtered_tb) from None
     68     finally:
     69       del filtered_tb

/usr/local/lib/python3.8/dist-packages/keras/engine/training.py in tf__train_function(iterator)
     13                 try:
     14                     do_return = True
---> 15                     retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
     16                 except:
     17                     do_return = False

/tmp/ipykernel_1421/3580991076.py in train_step(self, data)
     47         fake_image_and_labels = tf.concat([generated_images, image_one_hot_labels], -1)
     48         real_image_and_labels = tf.concat([real_images, image_one_hot_labels], -1)
---> 49         combined_images = tf.concat(
     50             [fake_image_and_labels, real_image_and_labels], axis=0
     51         )

ValueError: in user code:

    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1021, in train_function  *
        return step_function(self, iterator)
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1010, in step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1000, in run_step  **
        outputs = model.train_step(data)
    File "/tmp/ipykernel_1421/3580991076.py", line 49, in train_step
        combined_images = tf.concat(

    ValueError: Dimension 2 in both shapes must be equal, but are 7 and 9. Shapes are [28,28,7] and [28,28,9]. for '{{node concat_3}} = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32](concat_1, concat_2, concat_3/axis)' with input shapes: [?,28,28,7], [?,28,28,9], [] and with computed input tensors: input[2] = <0>.
#3

The main reason for this error is that real_image_and_labels and fake_image_and_labels cannot be concatenated because they don’t have the same size. As the error states,

real_image_and_labels has the shape [28,28,9]
fake_image_and_labels has the shape [28,28,7]

Please try to fix fake_image_and_labels variable’s shape to match that of real_image_and_labels. Thank you