These are chat archives for beniz/deepdetect

9th
Jan 2017
Abhinav Guru
@abhiguru
Jan 09 2017 18:51
Updated comment on the predict call failed issue Ubuntu 16.10 beniz/deepdetect#119
Emmanuel Benazera
@beniz
Jan 09 2017 19:04
are you using centos ?
Abhinav Guru
@abhiguru
Jan 09 2017 19:04
Ubuntu 16.10 if that was directed at me
Emmanuel Benazera
@beniz
Jan 09 2017 19:05
oops sorry
Abhinav Guru
@abhiguru
Jan 09 2017 19:05
np
Emmanuel Benazera
@beniz
Jan 09 2017 19:06
try moving the resnets deploy to exactly deploy.prototxt
remove the 101 and 152, you should use different directories, one for every model
Abhinav Guru
@abhiguru
Jan 09 2017 19:11
the error seem to have change in response {"status":{"code":500,"msg":"InternalError","dd_code":1007,"dd_msg":"src/caffe/util/io.cpp:45 / Check failed (custom): (fd) != (-1)"}}
Is it required to clear all and retrain after dir changes ?
*reload
Emmanuel Benazera
@beniz
Jan 09 2017 19:16
to recreate service yes. Make sure you have one caffemodel, the deploy and corresp files
if you are unsure try the exact steps of the tutorial
Abhinav Guru
@abhiguru
Jan 09 2017 19:35
it was the extra files in the folder i think which was causing the error didnt realize it was so picky
Is there a location to get the predict response cat codes too actual names of categories ? I get a response like {"status":{"code":200,"msg":"OK"},"head":{"method":"/predict","service":"imageserv","time":548.0},"body":{"predictions":[{"uri":"ambulance.jpg","classes":[{"prob":0.9999909400939941,"cat":"407"},{"prob":0.000007398507023026468,"cat":"734"},{"prob":6.831542691543291e-7,"last":true,"cat":"654"}]}]}}
Abhinav Guru
@abhiguru
Jan 09 2017 23:52
for #233 I keep getting failed to locate model template when only deploy.prototxt is present
but when the training prototxt is added to template folder it get created
is the value of nclasses 1000 ?