nilboy / colorization-tf

A Tensorflow implementation of ECCV2016 paper([Colorful Image Colorization](https://arxiv.org/pdf/1603.08511.pdf))

colorization-tf's Issues

Only call `softmax_cross_entropy_with_logits` with named arguments (labels=..., logits=..., ...)

i got an error
ValueError: Only call softmax_cross_entropy_with_logits with named arguments (labels=..., logits=..., ...)

dl2c = tf.stop_gradient(dl2c) - Failed to convert object of type <class 'list'> to Tensor

Traceback (most recent call last):
File "tools/train.py", line 23, in <module>
solver.train_model()
File "./solver.py", line 57, in train_model
new_loss, self.total_loss = self.construct_graph(scope)
File "./solver.py", line 45, in construct_graph
new_loss, g_loss = self.net.loss(scope, self.conv8_313, self.prior_boost_nongray, self.gt_ab_313)
File "./net.py", line 132, in loss
dl2c = tf.stop_gradient(dl2c)
File "/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 5220, in stop_gradient
"StopGradient", input=input, name=name)
.
.
.
TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [None]. Consider casting elements to a supported type.

PS: It seems to me that this script was made using Python 2.7, which tensforflow version do you use?
Could I ask for the dependencies?

Testing using the pre-trained model

I would like to test your model without having to train it first. I see that there is a color_model.ckpt file provided, but there is meta file with the graph. Is is possible to provide color_model.ckpt.meta as well?

I would like to ask can the training model handle any size image?

I would like to ask can the training model handle any size image?

tf.mul deprecated

tf.mul no longer exists (and has been renamed to tf.multiply: https://www.tensorflow.org/api_docs/python/tf/multiply)

tensorflow version

Could you show me which version of tensorflow you use?