Emmanuel Benazera

You are just need to fix you LD path, look the main README up.

@rdoume you shouldn't need to move file around, the build is automated but for the few TF Y/n questions. If you can lay out the exact thing you do (e.g....

Caffe and TF backend are optimized for CPU, with parallel operations using all cores. It is always best to use batches.

Hmm, this is used by `cppnetlib`. First I'd recommend to compile `cppnetlib` by hand as packaged versions are most often pretty old. Look ate the Ubuntu 15.10 and 16.04 deepdetect...

I think the server tells you what the problem is: the server is looking locally within the container for your model repo but my guess is that it is fact...

Also I'd recommend to keep the training and model repositories different

You need at least gcc4.8. You shouldn't need to tweak the app flags.

OK, two things: - try adding `#include ` to the headers of `imginputfileconn.h` - if that doesn't work, do you have any way to test with gcc5. I haven't tested...

OK, see PR #211 for the header fixes. This does not fix your `utfcpp` issue.

Looks like https://github.com/performous/performous/issues/201. Could you try against 0.11.2 at the moment ? Due to #207 users are testing the forthcoming 0.13 and they successfully built againt it on Ubuntu 16.04....