search

lucasb-eyer / pydensecrf

Python wrapper to Philipp Krähenbühl's dense (fully connected) CRFs with gaussian edge potentials.
MIT License
1.94k stars 414 forks source link

Training #7

Closed prassanna-ravishankar closed 8 years ago

prassanna-ravishankar commented 8 years ago

If I were to introduce a pure pythonic way of training with some kind of grid-search, instead of interfacing with the C code, do you think it would make sense?

To get : sx, sy, sr, sg, sb for the gaussian and bilateral functions

I'd put in a bunch of parameters, and see for which the error over my dataset is the least, and use those at test time. Does it make sense, or do I have to do more?

lucasb-eyer commented 8 years ago

I think that would be an OK thing to do in general, but I don't think this has a place in the wrapper: it belongs in your code. So I'm closing this.

If you need the wrapper to expose some additional functionality from the C library for this, feel free to re-open the issue and explain in more detail.