ECCV 2016
Face Hallucination
The source code is provided by https://github.com/XinYuANU/URDGN. Thanks for their works.
@inproceedings{Xin2016Ultra,
title={Ultra-Resolving Face Images by Discriminative Generative Networks},
author={Xin, Yu and Porikli, Fatih},
booktitle={European Conference on Computer Vision},
year={2016},
}
#52 x = create_YTC(pathfolder, 18) the size of low-resolution images
#61 x = create_YTC(pathfolder, 144) the size of high-resolution images
just change the code of train_ytc_xin_128_D.lua
#39 --scale (default 144) the size of high-resolution images
#48 ntrain = 1600 the num of train images
#49 nval = 40 the num of test images
#112 model_D:add(nn.Reshape(9 *9 *96)) --9 = 144 / (2 ^ 4)
#113 model_D:add(nn.Linear(9 *9 *96, 1024))
#198 local noise_inputs = torch.Tensor(N, 3, 18, 18)
#204 noise_inputs[{{i}}] = image.scale(torch.squeeze(noise_input_high[{{idx}}]),18,18)
#224 local to_plot = getSamples(valData_HR, 40) the num of test images
#232 local formatted = image.toDisplayTensor({input=to_plot, nrow=10})
#239 IDX = torch.randperm(1600) the num of train images
just change the code of adverserial_xin_v1_D.lua
#135 local LR_inputs = torch.Tensor(opt.batchSize, 3, 18, 18) the size of low-resolution images
#243 local sample = torch.Tensor(dataBatchSize, 3, 18, 18) the size of low-resolution images
#348 local inputs_lr = torch.Tensor(opt.batchSize, 3, 18, 18) the size of low-resolution images