CGAN之deepcolor实践 http://blog.csdn.net/kwame211/article/details/78203862?locationNum=4&fps=1
神经网络2. epoch, iteration, batchsize相关理解和说明
百行代码构建神经网络黑白图片自动上色系统
阅读文章:http://kvfrans.com/coloring-and-shading-line-art-automatically-through-conditional-gans/
尝试演示:http://color.kvfrans.com
安装程序
前提
Python 2.7,NumPy
- Tensorflow 0.12
OpenCV
运行它
1、make a folder called "results" 制作一个名为“results”的文件夹
2、make a folder called "imgs‘’ 做一个文件夹名为“imgs”
3、Fill the "imgs" folder with your own .jpg images, or run "download_images.py" to download from Safebooru.
填补“imgs”文件夹你自己的JPG图像,或者运行“download_images .py”下载来自Safebooru。
4、Run "python main.py train". I trained for ~20 epochs, taking about 16 hours on one GPU.
运行“ python main.py train”. 我训练了20个纪元,花了大约16个小时在一个GPU上。5、To sample, run "python main.py sample"
样本,运行“python main.py样本
6、To start the server, run "python server.py". It will host on port 8000.
要启动服务器,请运行“python server.py”。它将在8000端口上主机。
Pre-trained
前期的培训得到的pretrained模型:
https://drive.google.com/file/d/0BydPPLNieijIdDlUYWxhelEwRnM/view?usp=sharing
文件夹结构应该走
main.py
server.py
checkpoint/
tr/
checkpoint
model-1101500.index
model-1101500.data-00000-of-00001
model-1101500.meta