Training on customized dataset

[fangyan93]

@sunliamm Hello, I watched your video on YouTube which is really cool, and want to have a try to train the model on some other datasets. Could you please provide a general pipeline of training? Thank you very much!

[kevinlisun]

Hi Thanks for your interest! I am in busy time and have papers deadline due in mid Sep. I will release the full training pipeline by end Sep. Will update you once I have done it :)

[fangyan93]

@sunliamm Sure! Thank you very much! Looking forward to it! 😁

[fangyan93]

@ Hello, Mr. Sun, sorry to bother again. I am currently trying to train a 2 channel-model on nyud2 image dataset and have question about data preprocessing. In the paper, the RGB image and Depth image goes through 2 different network and then are combined together. I am wondering if the depth image used here is the direct output from Kinect, or is the depth data after preprocessing. Also, did you use any matlab toolbox in the real time test? Thank you very much!

Fangyan

[SCLiao47]

@sunliamm Hello, I am also curious about how to train on a customized dataset.
I have been trying to understand your code in "/dcnns" folders. But I am not really familiar with Caffe, so I can't figure it out.
I am wondering about why there is not a model saving line to save the trained caffemodel?
Also, I did not find out where you set the training directory.