These are chat archives for beniz/deepdetect

23rd
Sep 2016
dgtlmoon
@dgtlmoon
Sep 23 2016 07:01
ERROR - 02:17:19 - {"code":500,"msg":"InternalError","dd_code":1007,"dd_msg":"src/caffe/data_transformer.cpp:177 / Check failed (custom): (datum_height) == (height)"} I'm assuming that means one of my images was not atleast the dimensions as defined in the input { width: 500, height : 500 } ? Checking them all with file tells me they are all 500x500
Emmanuel Benazera
@beniz
Sep 23 2016 07:19
hi there, width and height are the input size of the network, the size of the images does not matter. It's usual to be 256x256 or 224x224.
dgtlmoon
@dgtlmoon
Sep 23 2016 07:21
x) I need todo some more tutorials i think
hmm, so i wonder what caused the error then?
rerunning @ 256x256 input network size
dgtlmoon
@dgtlmoon
Sep 23 2016 08:11
INFO - 08:06:09 - Solver scaffolding done. INFO - 08:09:42 - Ignoring source layer data INFO - 08:09:42 - Ignoring source layer label_data_1_split [08:09:42] /opt/deepdetect/src/caffelib.cc:1871: Error while filling up network for testing INFO - 08:09:42 - Opened lmdb tss-model/test.lmdb ERROR - 08:09:42 - service imageserv training status call failed ERROR - 08:09:42 - {"code":500,"msg":"InternalError","dd_code":1007,"dd_msg":"src/caffe/data_transformer.cpp:177 / Check failed (custom): (datum_height) == (height)"} Hmm something i'm not fully grocking here I think
happens after quite a few iterations
dgtlmoon
@dgtlmoon
Sep 23 2016 08:16
might try convnet template insteaed
Emmanuel Benazera
@beniz
Sep 23 2016 08:17
slowly follow the tutorial, it will work right out of the box. Custom convnet are more involved.