search

szq0214 / DSOD

DSOD: Learning Deeply Supervised Object Detectors from Scratch. In ICCV 2017.
Other
706 stars 210 forks source link

About 1 Channel images #29

Open didulle opened 6 years ago

didulle commented 6 years ago

Hi, I would like to learn from my own dataset composed of only gray level images. Could you tell me how I could adapt DSOD to work using only 1 channel. Thanks !!!

szq0214 commented 6 years ago

Hi @didulle, since DSOD is completely trained from scratch, I think the simplest way is to duplicate your channel twice and generate a new three-channel image as the input. You don't need to modify codes anymore. Of course, you can also change the data layer to fit your input format.

didulle commented 6 years ago

Thanks!! I have used the grayscale images as color ones and the results are good. I would like to check what's the best use 1 channel images or a color images constructed from gray level images. Could you advice how to change the data layer to fit 1 channel input image