Open EpsAvlc opened 4 years ago
Your question is too big to be answered in details (unless other members will have an urge to do it). But in a short, csv files contain voxelized data, features extracted from those voxels (e.g. curvature of the surface) and corresponding labels (cars, trees, ect.) plus some miscellaneous info. I'd advice you to take a look at notebooks section of our code, where we perform refinement and visualization of the data provided by Segmap. Probably, you can infer something from there.
Also, somewhere in the notebooks there are ids for N-ways classification, which are extremely important for the network to work properly. Don't miss them.
I want to write a pytorch version of this autoencoder. But I seldom could find the explanation of the dataset. Could you please explain each CSV file's usage ?