TorontoDeepLearning / convnet

A GPU implementation of Convolutional Neural Nets in C++
BSD 2-Clause "Simplified" License
504 stars 229 forks source link

fix type issue #40

Open avdmitry opened 8 years ago

avdmitry commented 8 years ago

Hello, Nitish @nitishsrivastava

I was inaccurate with types here, imread() reads image in the format CV_8UC3, which means 3 channels unsigned char. This code works with my tests when I implemented it, so I missed it, but recently I fixed this in my code, there it leads to incorrect behaviour. I didn't test much it here now, but anyway I should point out on possible issue. Sorry if this type mismatch leads to incorrect work of code.

I also made minor updates in the install document.

WBR, Dmitriy