During running of inference on NCS stick I got this error:
E:  dispatcherEventReceive:235 dispatcherEventReceive() Read failed -1
E: [ 0] eventReader:253 Failed to receive event, the device may have reset
E: [ 0] ncFifoDestroy:2525 Failed to write to fifo before deleting it!
I am using Neural Compute Stick with Raspberry pi 3 B+ connected using USB extension cable (due to NCS dimensions since it's too big to use it with other USB peripherals). The extension cable is not one of the cheapest (https://at.hama.com/00054504/hama-usb-30-extension-cable-shielded-050-m).
Can this cause a problem? How to make it work properly but avoiding directly plugging?
Thank you for reaching out, I have not tried this specific USB extension cable. I've connected the Neural Compute Stick to a Raspberry Pi using a powered USB hub without any issues.
- Are you running one of our sample codes?
- Are you using the Neural Compute Stick 1 or 2?
- Which SDK version are you using?
Hi @Jesus_at_Intel ,
Thanks for your feedback.
- I am testing my own MobileNet SSD classification, but of course using well established code snippets. I have compiled graph using caffe network as input.
- NCS 1
- SDK version 2.08.01.02
What's the best way to handle this exception real-time? Which method fires it? This one maybe: graph.LoadTensor(resized_image.astype(numpy.float16), None) ?
Also, will it reconnect automatically, or I should catch this exception, and initialize device and graph again?
I apologize for the delay in my response. The exception may be due to the device not getting enough power, try using a powered USB hub. I believe it will not automatically reconnect and the device would need to be disconnected and reconnected to the system.
Have you tried your model with our OpenVINO toolkit? It is also compatible with the NCS 1 and Raspberry Pi. Take a look at the following guide: