StanfordVL/taskonomy

Turning Taskonomy into a Multitask Network

bpeck81 opened this issue · 0 comments

I would like to use a shared encoder network trunk and different decoders to make a multitask network using taskonomy as a framework.

  1. Can this be feasibly done using the taskonomy project? (the codebase is large and I want to be sure before I attempt to work with it)
  2. I don't see the scripts to generate configs and I'm wondering where those are, as well as code/experiments/ and code/notebooks/ which seem like they seem like they might be helpful in seeing how to train the network.

Thanks!