I am new to Neural Compute Stick2. I have few quires as listed below can you please clarify me on that.
- Does NCS Stick2 support SSD_Mobilenet_v1(coco dataset) pre-trained/customized models trained using TensorFlow? Suggested links should be helpful
- I am just wondering whether I can run NCSDK version1 examples on Intel Neural Compute Stick2? If this is possible please provide me suggested links to do it.
- While trying to setup NCS stick 2 on my MAC machine I get this error “c++: internal compiler error killed program cc1plus”. I followed the link "https://software.intel.com/en-us/neural-compute-stick/get-started" I get this error on step4 Testing the installation.
Thanks for reaching out!
- Yes, OpenVINO has support for SSD MobileNet V1/V2 COCO. Take a look at the _Supported Topologies_ section in the following link: https://software.intel.com/en-us/articles/OpenVINO-Using-TensorFlow. You will need to freeze the model and generate the IR file (.xml & .bin) using the model optimizer.
- The NCSDK v1 and v2 only support the original Neural Compute Stick. OpenVINO is the official SDK for the Intel Neural Compute Stick 2. Take a look at the Get Started Guide to setup the NCS 2.
- At this time, OpenVINO is supported on Windows 10, Ubuntu 16.04, CentOS, and Raspbian OS.
- In addition to what PINTO suggested, I would recommend taking a look at the object detection sample code included with OpenVINO. You will need to follow the Raspberry Pi installation instructions.
Hope this helps! For additional questions around the OpenVINO Toolkit please post your questions on the Computer Vision Forum.
Hi @Jesus_at_Intel ,
I just knew that NCSDK v2 has supported Tensorflow SSD models recently. But the Release Note has listed only three of them, e.g.
- SSD Inception v2
- SSD Mobilenet v1
- SSD Mobilenet v2
I'm wondering whether the
SSD Lite Mobilenet v2 is also included (as that supported by OpenVINO)?
(Sorry for the sort of off-topic question. I was searching in the forum and found this thread, and thought that maybe asking my quick question here instead of staring a new thread is more feasible.)