@jeff123 Hi, the NCS is designed to be used as a low power inference device on the edge. This means that you can accelerate embedded devices like the Raspberry Pi. Typical workflow is to take your model and compile it to a Movidius graph format then use the NCSDK API to interface with the device and deploy your model. Currently the NCSDK has support for Caffe 1 and has support for some Tensorflow models/ops. You can see some of the sample code at https://github.com/movidius/ncappzoo
