FSRCNN
所属分类:人工智能/神经网络/深度学习
开发工具:C/C++
文件大小:8KB
下载次数:16
上传日期:2018-11-23 23:16:20
上 传 者:
夏雪风雷
说明: caffe写的FSRCNN代码,汤晓鸥大神发表在ECCV2016的论文代码,欢迎各位下载
(The FSRCNN code written by caffe, the code of the article written by Tang Xiaoou, published in ECCV2016, welcome to download)
文件列表:
FSRCNN\data_aug.m (727, 2016-09-09)
FSRCNN\FSRCNN-s_mat.prototxt (2565, 2016-09-09)
FSRCNN\FSRCNN-s_net.prototxt (2791, 2016-09-06)
FSRCNN\FSRCNN-s_solver.prototxt (577, 2016-09-09)
FSRCNN\FSRCNN_mat.prototxt (4052, 2016-09-09)
FSRCNN\FSRCNN_net.prototxt (4278, 2016-09-06)
FSRCNN\FSRCNN_solver.prototxt (578, 2016-09-09)
FSRCNN\generate_test.m (1962, 2016-09-09)
FSRCNN\generate_train.m (2040, 2016-09-09)
FSRCNN\modcrop.m (279, 2015-03-17)
FSRCNN\plot_filterpattern.m (841, 2016-09-09)
FSRCNN\saveFilters.m (2391, 2016-09-09)
FSRCNN\store2hdf5.m (2897, 2015-06-05)
FSRCNN\test.txt (24, 2016-09-06)
FSRCNN\train.txt (25, 2016-09-06)
FSRCNN (0, 2018-11-02)
***********************************************************************************************************
Training code for "Accelerating the Super-Resolution Convolutional Neural Networks" (ECCV 2016)
by Chao Dong (dongchao@sensetime.com)
***********************************************************************************************************
Usage:
1. Place the "FSRCNN" folder into "($Caffe_Dir)/examples/"
2. Open MATLAB and direct to ($Caffe_Dir)/example/FSRCNN, run
"generate_train.m" and "generate_test.m" to generate training and test data.
You can also run data_aug.m to do data augmentation first.
3. To train our FSRCNN, run
./build/tools/caffe train --solver examples/FSRCNN/FSRCNN_solver.prototxt (For FSRCNN)
or
./build/tools/caffe train --solver examples/FSRCNN/FSRCNN-s_solver.prototxt (For FSRCNN-s)
4. After training, you can extract parameters from the caffe model and save them in the format that can be used in our test package (FSRCNN_test). To do this, you need to install mat-caffe first, then open MATLAB and direct to ($Caffe_Dir) and run "saveFilters.m". The "($Caffe_Dir)/examples/FSRCNN/FSRCNN.mat" will be there for you.
近期下载者:
相关文件:
收藏者: