Mingming He

@mnbrshd The algorithm and energy function is described in Sec 3.4. To find the optimization for minimizing the energy function is a Conditional Random Field problem that can be solved...

Please try [this solution](https://github.com/hmmlillian/Neural-Color-Transfer/issues/3#issuecomment-599141422).

@iperov Good suggestion! We should do that.

Hi @songyn95 , since our model is trained with grayscale images in ImageNet dataset, I do not think it can be directly applied to infrared images. You may need to...

@songyn95 I think both networks (vgg19_gray and colorization subnet) need to be retrained, since the input image of both networks have to be replaced by infrared images. To train a...

@AliAsafPolat I tested on the image with a resolution of 256x256 pixels on a notebook GTX 1070, the first step to compute bidirectional correspondences takes about 18 seconds while the...

Hi @SibingWu, we resized the images with their short edges of 256 pixels (while allowing their long edges to be any arbitrary length to keep aspect rations) during training. It...

@coallar please generate the bidirectional flow and combo file first following the two steps below: - A command to generate bidirectional mapping functions using *Deep Image Analogy*: ``` deep_image_analogy.exe [MODEL_DIR]...

> 您好， 我想请教一下关于test data那里，我们没有像训练数据reference匹配时的类别信息，不能按照类别去寻找reference，需要将每个图片与ImageNet中所有数据去对比 然后找最相近的数据嘛？ 您好，可以参考这个[repo](https://github.com/hmmlillian/Gray-Image-Retrieval)的实现先找到对应的类，然后再在类里面找最接近的图片。

@xuzhongyou Yes, that's a problem. We have to do some modification before compiling the code on Linux. We will do it soon. Thanks for your suggestion.