These are chat archives for beniz/deepdetect

29th
Apr 2016
tmeroniown
@tmeroniown
Apr 29 2016 14:00
Hello - I installed deepdetect on an amazon AMI. How can I access server logs?
Emmanuel Benazera
@beniz
Apr 29 2016 14:00
hi @tmeroniown are you using a docker container or a straight install from source ?
tmeroniown
@tmeroniown
Apr 29 2016 14:01
hi @beniz . I'm not using a container - I installed everything from scratch.
Emmanuel Benazera
@beniz
Apr 29 2016 14:01
then server logs are the output of dede, in place from where you start it
tmeroniown
@tmeroniown
Apr 29 2016 14:01
so just whatever I see on the shell?
Emmanuel Benazera
@beniz
Apr 29 2016 14:03
that's correct, at the moment, it is for you to redirect it wherever you want to consume it
tmeroniown
@tmeroniown
Apr 29 2016 14:03
very good. thank you!
btw - it took me forever to figure out that I need to start ./dede with the -host 0.0.0.0 flag
in order to access it from outside that is
If you add that to the documentation i think it would be helpful for others trying to do it
Emmanuel Benazera
@beniz
Apr 29 2016 14:06
like ./dede --help ? Also here: http://www.deepdetect.com/overview/usage/ No criticism intended, but then let us know where to put it clearer, in the README.md ?
tmeroniown
@tmeroniown
Apr 29 2016 14:07
sorry - I did not mean to criticize :-)
the biggest problem in the docs today is that all examples assume localhost and 8080
it was actually a bit tricky to get everything to work when those assumptions are broken
well - at least it was a bit tricky for me :-)
Emmanuel Benazera
@beniz
Apr 29 2016 14:08
OK, got it. Thanks. Would you like to add an example to the README.md via a PR ?
tmeroniown
@tmeroniown
Apr 29 2016 14:11
Sure I'd love to help. I will do that. Thanks a lot for everything!
Emmanuel Benazera
@beniz
Apr 29 2016 14:12
thanks, I believe it is actually a good practice to leave part of the doc into the hands of users since it can fix difficulties in the right places.
tmeroniown
@tmeroniown
Apr 29 2016 14:12
yes I agree
Emmanuel Benazera
@beniz
Apr 29 2016 14:15
Other improvements are welcome. Unfortunately at the moment the website tutorials and documentation are not in public git access though I would love to see the tutorials and examples pages open to contribution...
tmeroniown
@tmeroniown
Apr 29 2016 15:06
@beniz - question for you. I'm using the googlenet model to extract feature vectors. In standard caffe examples, they talk about things like subtracting the mean image, and all sorts of other pre-processing. is dede doing that?
Emmanuel Benazera
@beniz
Apr 29 2016 15:06
dede is doing that automatically if the mean.binaryproto file is in the model repository
tmeroniown
@tmeroniown
Apr 29 2016 15:09
I see. where do I get that file from?
Emmanuel Benazera
@beniz
Apr 29 2016 15:10
this should come along with the model, and when not, you may not want to bother too much.
tmeroniown
@tmeroniown
Apr 29 2016 15:10
also in googlenet there is a channel swap
Emmanuel Benazera
@beniz
Apr 29 2016 15:10
typically dd automatically produces that file when training a model
swap ?
tmeroniown
@tmeroniown
Apr 29 2016 15:14
scratch that - I think its only relevant if you are using the Pycaffe interface