Stuck at "Loading model to the plugin" when using NCSV2

[Tse__Ho_Cheong_Edwin]

Hi everyone.

I used faster-rcnn-inception-v2-coco to train my model and successfully transforming to IR using Openvino.

However, when I try to run the object_detection_sample_ssd using NCSV2, it stuck at here:

[ INFO ] Loading model to the plugin

I have added more info message to debug and finds out that it probably stuck at :

ExecutableNetwork executable_network = plugin.LoadNetwork(network, {});

I am using VM with Ubuntu and have no problem running demo_security_barrier_camera and demo_squeezenet_download_convert_run with NCSV2 but get stuck with my own model.

Thank you.

[Shubha_R_Intel]

How does it work outside of a VM environment ? You can try it on Windows 10 (it works).

[Tse__Ho_Cheong_Edwin]

No, it also stuck at "[ INFO ] Loading model to the plugin" when using Windows 10.

[Shubha_R_Intel]

Hello Ho Cheong. Here's a debugging technique. Can you kindly try with -d GPU first on Windows 10 ? (I assume you have an Intel GPU). Or if not GPU, try -d CPU. Let us know what happens.

[Tse__Ho_Cheong_Edwin]

Thank you for your reply.

I’ve tried to generate two xml with Datatype FP16 and FP32.

Since cpu doesn’t support FP16, I used FP32 with cpu and it works, a bmp image was successfully created without any problem.

But when I use NCSV2 with a FP16 xml, no matter I’m running on a Ubuntu VM or Windows 10, both stuck at loading model to the plugin.

It seems that the program is running since the NCSV2 become warmer, but I tried to let it run for 8 hours and still no hope.

Thank you.

[om77]

Hi,

it's very possible that your network reaches a some resource limitation for stick (cpu, memory).

To ensure if this is the case, try to reduce network inputs array.

E.g. in the case of faster-rcnn-inception-v2-coco

https://github.com/tensorflow/models/blob/master/research/object_detection/samples/configs/faster_rcnn_inception_v2_coco.config

try to reduce image_resizer parameters e.g. to 480-576.

[Tse__Ho_Cheong_Edwin]

Thank you om77.

I have reduced the  image_resizer parameters to 480-576 but still stuck at the [ INFO ] Loading model to the plugin.

It is a little bit strange because it works perfectly on CPU.

Here is the log info:

C:\Users\XXXXXXXX\Documents\intel\OpenVINO\inference_engine_samples_2017\intel64\Debug>object_detection_sample_ssd.exe

-i "c:\Users\XXXXXXXX\Desktop\photo\000241.jpg"

-m "c:\Users\XXXXXXXX\Desktop\inference\inference_smaller_16\frozen_inference_graph.xml" -d MYRIAD

[ INFO ] InferenceEngine:
        API version ............ 1.4
        Build .................. 19154
Parsing input parameters
[ INFO ] Files were added: 1
[ INFO ]     c:\Users\XXXXXXXX\Desktop\photo\000241.jpg
[ INFO ] Loading plugin

        API version ............ 1.5
        Build .................. 19154
        Description ....... myriadPlugin
[ INFO ] Loading network files:
        c:\Users\XXXXXXXX\Desktop\inference\inference_smaller_16\frozen_inference_graph.xml
        c:\Users\XXXXXXXX\Desktop\inference\inference_smaller_16\frozen_inference_graph.bin
[ INFO ] Preparing input blobs
[ INFO ] Batch size is 1
[ INFO ] Preparing output blobs
[ INFO ] Loading model to the plugin

I also tried to downgrade from R5.01 to R5 but still get stuck at the same place.

Thank you.

[nikos1]

Hello Tse, Ho Cheong Edwin

Could you try the following command so that we can debug the issue further

Copy contents including sub-folders of C:\Intel\computer_vision_sdk_2018.5.456\deployment_tools\inference_engine\bin\intel64\Debug 

over to C:\Users\XXXXXXXX\Documents\intel\OpenVINO\inference_engine_samples_2017\intel64\Debug

and then open command prompt in C:\Users\XXXXXXXX\Documents\intel\OpenVINO\inference_engine_samples_2017\intel64\Debug and run from there - NCS still stuck?

If that fails for Debug try same steps for Release

BTW are you using FP32 or FP16 for NCS ?

Cheers,

nikos

[whatthisismyname]

Did you find the solution to this? i change the image_size too but the problem persisted?