Yijun Li

@binbinmeng I modified the `test.py` by adding `volatile=True` and find that it can test large image size (e.g., 1024) now. Have a try.

@suke27 Sorry for the late reply. It is not a must and you can just input your original image. Sometimes I found that style patterns (e.g., if the style image...

@Tobias-Fischer Thanks for your interests. We have not planned to do that. However, we heavily borrow the code from this amazing [SegTool](http://www.cs.cmu.edu/~mohitg/segmentation.htm) UI demo, which can be a good sample...

@wzm2256 Thanks for your interests about our work. You can set values for two bits (e.g., 0.5 for each bit) in the selection unit to generate mixed texture (Line 48...

@swordcheng Thanks for your interests. You can take a look at Section 3.5 of the [paper](http://openaccess.thecvf.com/content_cvpr_2017/papers/Li_Generative_Face_Completion_CVPR_2017_paper.pdf) where I present the three-stage training strategy. There might be some failures as the...

You can adjust the loss weight (`0` or not) to separate the training into different stages.

@shz1314 Thanks for your interests. All network architectures and training parameters can be found [here](https://github.com/Yijunmaverick/GenerativeFaceCompletion/tree/master/matlab/FaceCompletion_training/model). By the way, you can refer this tensorflow [implementation](https://github.com/jazzsaxmafia/Inpainting) of a similar paper [ContextEncoder](https://github.com/pathak22/context-encoder).

@taki0112 Sorry, this is not my work. You can ask authors for details.

@taki0112 Check the author's reply above

@okdewit @Multiboxer Thanks for your suggestions on estimating the memory. High-resolution is always a challenging issue in deep models. To run my code on GPUs with small memory, you need...