Open chrod opened 6 years ago
@chord, first of all thanks for debugging the YOLO model.
Can you give me some information on your system, e.g.:
I checked the models with the last Windows release of Caffe (GPU) under Python 3.5 (Anaconda) and it's working correctly. I'll also check on my Linux laptop to verify that the layer parameters didn't change recently.
I'm also assuming that you are using the converted .caffemodel
files from YOLO; the links are at the end of the Installation section.
The
tiny yolo
andtiny yolo voc
caffemodel
files don't seem to agree with theprototxt
files.I get an error that the source layer blob size mismatches for
conv1_bn
. (see below)Trying to troubleshoot didn't completely fix the problem (Changed the
bias_term
value totrue
in conv1_scale, and ran into another similar error for layerconv1
(bottom))If I'm on the right track in debugging this, do you have updated matching prototxt files to match the caffemodel files?
Thanks, Chris
Error#1 (conv1_bn) text:
Error#2 (conv1) text: