zlingkang / mobilenet_ssd_pedestrian_detection Goto Github PK

I tried your model on my code to see if my pi can go faster as its taking like 1.3 frames per second but i got same performance but less tracking any reasons why?

When trying to run camdemo.py I got this error

WARNING: Logging before InitGoogleLogging() is written to STDERR
W1218 13:23:20.708622 20025 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W1218 13:23:20.708693 20025 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W1218 13:23:20.708703 20025 _caffe.cpp:142] Net('MobileNetSSD_deploy.prototxt', 1, weights='MobileNetSSD_deploy10695.caffemodel')
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format caffe.NetParameter: 1177:17: Message type "caffe.LayerParameter" has no field named "permute_param".
F1218 13:23:20.716996 20025 upgrade_proto.cpp:88] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: MobileNetSSD_deploy.prototxt
*** Check failure stack trace: ***

I'm using Caffe pre-compiled from this repository https://launchpad.net/ubuntu/+source/caffe
Can you tell me which version of Caffe should I use
Thank you very much!

Hi, zlingkang.

I have a strange problem when dealing with the Caltech dataset. Bascially, I think we are using the same dataset. I also collect all the pedestrian data from the originall Caltech dataset, and I also changed the name from pedestrian to person.

But, when I start to train the dataset, I get the loss keeps being 0.

Like this issue I create before.
chuanqi305/MobileNet-SSD#116

Have you met that kind of situation? And Could you share your solver_train.prototxt, I think maybe I got some wrong Hyper parameters setting in this file.

• In case you are Chinese, you can reply directly in Mandarin.

Many Thanks , Looking forward for your reply.

How's your log-average miss rate in Caltech?