Detection of bounding box for color images

[mra...@email.wm.edu]

Hi,

My goal is to train a convnet for detecting road sign. Train images are 28x28 and all of them are color images. At the detection phase I want to detect the bounding boxes and to save it. I am using the configuration file used for mnist demo. But, mnist is not for only one input channel. So, what should I change in the mnist configuration file to work for RGB images and to detect the bounding boxes?

I am very new in convolutional neural network and EBlearn. So, pardon me for my ignorance.

Thanks
Atiqur