search

leehomyc / Photo-Realistic-Super-Resoluton

Torch Implementation of "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"
MIT License
203 stars 58 forks source link

test image folder can not be reused #13

Open hannraoi opened 7 years ago

hannraoi commented 7 years ago

Hi~it's me again and this time I have met another problem =)

The first time I run "t_folder=/xx/xx model_file=/xx/xx result_path=/xx/xx th run_test.lua" and it worked perfectly, which is nice, but after I change the test images under the same folder, it reminds me that a certain image file can not be found, the error info is like this:

hannraoi@hannraoi-B85M-DS3H:~/Photo-Realistic-Super-Resoluton-master$ t_folder=/home/hannraoi/my_srgan_test/ model_file=/home/hannraoi/my_srgan_model/super_resolution_adversarial_G_134 result_path=/home/hannraoi/my_srgan_result/ th run_test.lua { niter : 250 dataset : "folder" ntrain : inf model_file : "/home/hannraoi/my_srgan_model/super_resolution_adversarial_G_134" t_folder : "/home/hannraoi/my_srgan_test/" batchSize : 32 result_path : "/home/hannraoi/my_srgan_result/" loadSize : 96 nThreads : 4 scale : 4 gpu : 1 } Starting donkey with id: 1 seed: 1 table: 0x7fd2180ff0a0 Starting donkey with id: 3 seed: 3 table: 0x7fd214129350 Starting donkey with id: 2 seed: 2 table: 0x7fd210128ad0 Starting donkey with id: 4 seed: 4 table: 0x7fd20c128da0 Loading train metadata from cache Loading train metadata from cache Loading train metadata from cache Loading train metadata from cache /home/hannraoi/torch/install/bin/lua: ...hannraoi/torch/install/share/lua/5.2/threads/threads.lua:183: [thread 2 callback] /home/hannraoi/torch/install/share/lua/5.2/image/init.lua:352: /home/hannraoi/my_srgan_test/subfolder/1b38b72e-af9c-4708-8e7d-5ba80d4a24ea.png: No such file or directory stack traceback: [C]: in function 'error' /home/hannraoi/torch/install/share/lua/5.2/image/init.lua:352: in function 'load' ...tic-Super-Resoluton-master/data/donkey_folder_supres.lua:35: in function 'loadImage' ...tic-Super-Resoluton-master/data/donkey_folder_supres.lua:50: in function <...tic-Super-Resoluton-master/data/donkey_folder_supres.lua:48> (...tail calls...) .../Photo-Realistic-Super-Resoluton-master/data/dataset.lua:354: in function <.../Photo-Realistic-Super-Resoluton-master/data/dataset.lua:347> (...tail calls...) [C]: in function 'xpcall' ...hannraoi/torch/install/share/lua/5.2/threads/threads.lua:234: in function 'callback' ...e/hannraoi/torch/install/share/lua/5.2/threads/queue.lua:65: in function <...e/hannraoi/torch/install/share/lua/5.2/threads/queue.lua:41> [C]: in function 'pcall' ...e/hannraoi/torch/install/share/lua/5.2/threads/queue.lua:40: in function 'dojob' [string " local Queue = require 'threads.queue'..."]:15: in main chunk stack traceback: [C]: in function 'error' ...hannraoi/torch/install/share/lua/5.2/threads/threads.lua:183: in function 'dojob' ...aoi/Photo-Realistic-Super-Resoluton-master/data/data.lua:76: in function 'getBatch' run_test.lua:33: in main chunk [C]: in function 'dofile' ...raoi/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk [C]: in ?

I know little about Lua, so I don't know where the problem comes from, I hope to get some help, thanks!

hannraoi commented 7 years ago

by the way, my solution to the problem is to create a new t_folder with a different name each time I run the test code.

smahajan07 commented 7 years ago

Hi @hannraoi . How many images did you use from Image net to train the model? And which data did you test your model on?

I have trained on a data of faces (approx 200,000) till 50 epochs, but the results are not great.

hannraoi commented 7 years ago

hi~ @smahajan07 I'm sorry that I didn't use ImageNet data as my training data, I use my own data instead and the amount of data is a little less than urs.