Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.
6404 Discussions

looking for a caffe example in the SDK, or otherwise, that uses caffe::caffe_copy() to the NCS

idata
Employee
664 Views

Hello,

 

I am not convinced the ~/Movidius/ncsdk-1.10.00.03/ncsdk-1.10.00.03/examples/caffe actually use the caffe::caffe_copy() or other caffe APIs to pass and receive data from the NCS gpu.

 

can you direct me to an example of this that actually uses the NCS SDK implementation of the caffe APIs?

 

Thanks,

 

Mark
0 Kudos
3 Replies
idata
Employee
470 Views

@mdouglas_95136, NC SDK uses Caffe framework only to parse the prototxt file and compare inference results when you run mvNCChecker. NC SDK API does note require Caffe APIs - all it does is load the binary graph file, load images for inference and read out the results. Refer the following for more detailed explanation:

 

0 Kudos
idata
Employee
470 Views

@AshwinVijayakumar thanks for the details.

 

--- Switching gears and looking at the Movidius implementation of caffe support in terms of the caffe APIs (i.e. caffe::caffe_copy()) and their respective methods, can you tell me if these have been tested? and to what extent?

 

--- I ask because I will be calling the caffe_copy() and other caffe APIs directly from a user-space application inside a Docker container next week.

 

--- Can you tell me at least that your caffe APIs/methods have been tested to some extent?

 

Thanks,

 

Mark D
0 Kudos
idata
Employee
470 Views

Mark, Please look at the following links for high level understanding of how NCS works.

 

https://movidius.github.io/ncsdk/ncs1arch.html

 

https://movidius.github.io/blog/qconf-sfo/

 

Also, our full Caffe layer support is documented here https://movidius.github.io/ncsdk/Caffe.html

 

If that does not answer what you are looking for, please share your prototext file with us and we can help you further.

0 Kudos
Reply