Closed Liming-Cheng closed 3 years ago
Recently I have been dabbling in PyTorch Python training code. I think, in general, the inference code should be very similar to what I did in Monodepth-cpp: 1. load the network, 2. define input and output layers, 3. feed the input, and 4. get the output. Unfortunately, I am quite busy with my research lately. However, I will keep this issue open so that other people can come in and get in touch with you.
Hi ! yan, recently I found an efficent work like monodepth, and unfortunately it was implemented by python and pytorch. So it cannot be used in slam easily. Here is the link: https://github.com/JiawangBian/SC-SfMLearner-Release To implement it by c++, libtorcch must be compiled and the model .th be converted to .pt, and all of this maybe simple, but what is difficult is to convert the python code to c++.
You have implemented monodepth by c++ , so lt will be a great contribution to this filed if you have interest in implementing the work in c++. I hope sincerely you can consider it when you are not busy , thank you! Good luck for everything!