Segmentation fault

[diego-varela-seijas]

Hi,

I need to do inference on a Raspberry Pi 4B with NCS 2 using YOLO. I get the default model in IR (the xml and the bin file) before converting custom models to try, and used two demos of open_model_zoo but I always get a Segmentation fault error. Here are the outputs:

Command
./multi_channel_object_detection_demo_yolov3 -m ../../../build/frozen_darknet_yolov3_model.xml -d MYRIAD -i ../../../build/cars.mp4

Output:
[ INFO ] InferenceEngine: IE version ......... 2021.4 Build ........... 0 [ INFO ] Parsing input parameters [ INFO ] Detection model: ../../../build/frozen_darknet_yolov3_model.xml [ INFO ] Detection threshold: 0.5 [ INFO ] Utilizing device: MYRIAD [ INFO ] Batch size: 1 [ INFO ] Number of infer requests: 5 [ INFO ] Model path: ../../../build/frozen_darknet_yolov3_model.xml Segmentation fault

Command
./object_detection_demo -m ../../../build/frozen_darknet_yolov3_model.xml -d MYRIAD -i ../../../build/cars.mp4 -at yolo

Output:
[ INFO ] InferenceEngine: IE version ......... 2021.4 Build ........... 0 [ INFO ] Parsing input parameters [ INFO ] Reading input [ WARN:0] global ../opencv/modules/videoio/src/cap_gstreamer.cpp (1044) open OpenCV | GStreamer warning: unable to query duration of stream [ WARN:0] global ../opencv/modules/videoio/src/cap_gstreamer.cpp (1081) open OpenCV | GStreamer warning: Cannot query video position: status=1, value=0, duration=-1 [ INFO ] Loading Inference Engine [ INFO ] Device info: [ INFO ] MYRIAD myriadPlugin version ......... 2021.4 Build ........... 0 Loading network files [ INFO ] Batch size is forced to 1. [ INFO ] Checking that the inputs are as the demo expects [ INFO ] Checking that the outputs are as the demo expects [ INFO ] Loading model to the device Segmentation fault

I don't know what could be the error or how to fix it. If anyone knows another way to do inference with YOLO on a Raspberry with the NCS it would be also good.

Thanks in advance

[vladimir-dudnik]

@alvoron @terfendail could you please look at the issue?

[alvoron]

@diego-varela-seijas thank you for raising the issue. We will try to reproduce it at the nearest time.
The cause of segmentation fault is often related to model size. Did you consider to try tiny version of YOLO?

[diego-varela-seijas]

The version was already YOLOv3 tiny

[alvoron]

@diego-varela-seijas did you build OMZ demo multi_channel_object_detection_demo_yolov3 on Raspberry Pi?

[diego-varela-seijas]

Yes, the only step I didn't do on RPi was converting the model to .xml and .bin.

[alvoron]

@diego-varela-seijas could you try to add -DCMAKE_CXX_FLAGS="-march=armv7-a" to cmake command when you build OMZ demo?

[diego-varela-seijas]

Adding that made it work. Thank you!!