Movidius NCS? Check!
RPi with Movidius SDK? Check!
A trained object detection model? Check!
Scripts that successfully run inference on the RPi using the NCS? Check!
Compatibility between my model and the Movidius SDK installed version of caffe?…
Nope!
The issue that I'm facing is that the model I have trained is Caffe fast-rcnn - and includes the ROI_Pooling_Layer (Layer 7) which isn't present in the SSD flavor of Caffe that the Movidius SDK uses.
My solution is to find a way to include the rcnn branch of Caffe instead. If anyone has any advice on this, 't'would be much appreciated!
In the meantime, I'm gonna plug away and try to fix it.
