Open tanakka opened 8 years ago
@tanakka FYI #16 I post a piece of code in the reply.
I know pre-computed proposals are given in the torch format since the file is huge and hard to create it at our end, but is there a straightforward approach to create these files when your dataset is small. Like @tanakka asked, whats the next step in converting those split json or t7 files into train.t7 or val.t7, is there a code snippet to make this process easier? Thanks in advance.
How can I generate train.t7 and val.t7 from deepmask output? deepmask/evalPerImage.lua generates multiple JSON|t7 files. (divided into chunks of 500 images each)