jzbontar / mc-cnn

Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches
BSD 2-Clause "Simplified" License
707 stars 232 forks source link

Training based on the pre-trained net using new data #54

Open GoodStudyDayUpUp opened 6 years ago

GoodStudyDayUpUp commented 6 years ago

@jzbontar Hello, thanks a lot for sharing the code. That's amazing. Now I want to train a net based on your pre-trained net using some new data, which means I want to start from the pre-trained neural network (e.g. net_mbslow-a_train_all.t7) instead of from the beginning using an empty net. And I only want to train the first half of the net because I only want to teach the net to extract feature from my data. I will appreciate a lot if you could share some suggestions! Looking forward to your reply! Best regards,