Liu Yang

@sniklaus It's a great work about video frame interpolation! I'm trying to implement it by myself. While I am a little confused about some details of your spatial warping. -...

@turboLIU I trained the model only by DIV2K training data. While the pretrained_model in checkpoint was trained with the option `--matlab_bicubic=True`. There is some difference in the implement of bicubic...

@yumingcong 您好，如果训练就按照你的理解建立文件夹,把数据放进去，如果简单测试的话，工程里有训练好的模型，可以直接运行`python main.py --is_train=False --matlab_bicubic=True`

@difcreate 我也不确定这个问题原因是什么，你可以用最新的代码试试

@13135675192 Hi, did you run the code on windows? I have modified the code, but not tested it on windows yet. You could pull the latest commit and try it.

@13135675192 Did you test the pretrained model? It needs to be tested with `--matlab_bicubic=True`.

Never mind, “--matlab_bicubic=True” is not necessary but should be same when training and testing.

You could run `python main.py` or `python main.py --matlab_bicubic=True`. While from my observation, setting "--matlab_bicubic=True" leads to better performance.

@MercedesMT Did you download the DIV2K data set and put all images to RDN-TensorFlow/Train/DIV2K_train_HR？

@rzumer The phase shifting function is a third implement of the pixel shuffle layer in [ESPCN](https://arxiv.org/abs/1609.05158). Theoretically it does not change the tensor if scale=1. You could try commenting out...