search

vqnhat / DSN-Binarization

Multi-scale DSNs for degraded document image binarization
11 stars 3 forks source link

can not load image #3

Closed justinner closed 5 years ago

justinner commented 5 years ago

Hi,I am sorry to bother you again,I have downloaded your total source,and I want to train,but I met the problem ,

/../data/binarization/train_pair_for_2011.lst I1214 11:25:34.042053 6563 image_labelmap_data_layer.cpp:50] Shuffling data I1214 11:25:34.060681 6563 image_labelmap_data_layer.cpp:55] A total of 84210 images. E1214 11:25:34.060843 6563 io.cpp:77] Could not open or find file ../../../data/binarization/train_data\028923_45.png E1214 11:25:34.060851 6563 io.cpp:77] Could not open or find file ../../../data/binarization/train_data\028923_45_gt.png F1214 11:25:34.060855 6563 image_labelmap_data_layer.cpp:84] Check failed: cv_img.data Could not load train_data\028923_45.png Check failure stack trace: 已放弃 (核心已转储)

Even if I deleted the image,I still had the error but not the same image,could you help me?

justinner commented 5 years ago

hi,Vqnhat,I have run your code successfully,but when I try to modify the train_val_conv3.prototxt,the error was input + output channels need to be the same I add upsample layer,I want to ask if the error is important,or can I ignore

vqnhat commented 5 years ago

Hi,

About this problem, I guess you should check the output of your modified network. If there are no problem with the generated images, you can ignore the error message.

justinner commented 5 years ago

OK,thanks very much,I will try again.