Intel® DevCloud
Help for those needing help starting or connecting to the Intel® DevCloud
1695 Discussions

Unable to use the yolov7-tiny in IR format to Intel-DevDcloud for the edge

Wen_Jie
Beginner
1,504 Views

Greeting to all, I hope you're all doing well. I have a question that I need help with, and I'm hoping to receive a prompt response. I appreciate your time and expertise in advance.

 

At first, I export the yolov7-tiny.pb to yolov7-tiny.onnx, then convert the yolov7-tiny.onnx to the IR format using openvino 2022.2.0.

 

After I got the IR format, I was trying to deploy yolov7-tiny to Intel-DevDcloud for the edge by modifying the code inside the example: "Reference-samples/iot-devcloud/openvino-dev-latest/developer-samples/python/tiny-yolo-v4-python" as shown in the figure below.

 

Wen_Jie_0-1681658371508.png

However, the inferencing fails and comes out with the error message (ValueError: get_shape was called on a descriptor::Tensor with dynamic shape) as shown in the figure below.

Wen_Jie_1-1681658500279.png

 

What should I do to overcome this problem?

I'm looking forward to hearing from you soon. Thank you!

I had attached the model files in the attachments section.

0 Kudos
1 Solution
JesusE_Intel
Moderator
1,357 Views

Hi Wen_Jie,

 

Since the model is running with Benchmark_app you should be able to use it and obtain inference results. You will need to write your own code as we don't have a sample/demo for yolov7-tiny.

 

If can also try to run your model with the benchmark_app tutorial that does not require an AWS account. You can find the tutorial in the following directory via JupyterLab.

Reference-samples/iot-devcloud/openvino-dev-latest/tutorials/benchmarkApp_python/

 

Simply upload your model to the model/FP16 directory and update the following cell.

JesusE_Intel_0-1682710128773.png

 

Regards,

Jesus

 

View solution in original post

0 Kudos
4 Replies
JesusE_Intel
Moderator
1,469 Views

Hi Wen_Jie,


Thanks for reaching out, have you tried running your model with the benchmark_app (Reference-samples/iot-devcloud/openvino-dev-latest/developer-samples/python/benchmarkApp-python/) to ensure inference is possible with OpenVINO?


That being said, we don't have an OpenVINO sample/demo available for YoloV7-tiny on Intel Developer Cloud for the Edge. You could try to implement your own code by referencing the object detection demo from Open Model Zoo. However, this demo code will also need to be modified to support YoloV7-tiny as it only supports the following architectures.


  • architecture_type = yolo
    • mobilefacedet-v1-mxnet
    • mobilenet-yolo-v4-syg
    • person-vehicle-bike-detection-crossroad-yolov3-1020
    • yolo-v1-tiny-tf
    • yolo-v2-ava-0001
    • yolo-v2-ava-sparse-35-0001
    • yolo-v2-ava-sparse-70-0001
    • yolo-v2-tf
    • yolo-v2-tiny-ava-0001
    • yolo-v2-tiny-ava-sparse-30-0001
    • yolo-v2-tiny-ava-sparse-60-0001
    • yolo-v2-tiny-tf
    • yolo-v2-tiny-vehicle-detection-0001
    • yolo-v3-tf
    • yolo-v3-tiny-tf
  • architecture_type = yolov3-onnx
    • yolo-v3-onnx
    • yolo-v3-tiny-onnx
  • architecture_type = yolov4
    • yolo-v4-tf
    • yolo-v4-tiny-tf
  • architecture_type = yolof
    • yolof
  • architecture_type = yolox
    • yolox-tiny

Ref: open_model_zoo/demos/object_detection_demo/python at master · openvinotoolkit/open_model_zoo · GitHub


Hope this helps!


Regards,

Jesus


0 Kudos
Wen_Jie
Beginner
1,451 Views

Hi Jesus

 

Could you show me which line of codes should I modify in order to run the model with the benchmark_app (Reference-samples/iot-devcloud/openvino-dev-latest/developer-samples/python/benchmarkApp-python/)?

I did not have any AWS, so I will need to upload the model manually.

In the meantime, I had try to run the benchmark_app in the Ubuntu 18.04 as shown below.Wen_Jie_0-1681787995724.png

Wen_Jie_1-1681788023266.png

It successfully benchmarking the yolov7-tiny in IR format.

So this means the inference can be run with OpenVINO for the yolov7-tiny in IR format?

 

0 Kudos
JesusE_Intel
Moderator
1,358 Views

Hi Wen_Jie,

 

Since the model is running with Benchmark_app you should be able to use it and obtain inference results. You will need to write your own code as we don't have a sample/demo for yolov7-tiny.

 

If can also try to run your model with the benchmark_app tutorial that does not require an AWS account. You can find the tutorial in the following directory via JupyterLab.

Reference-samples/iot-devcloud/openvino-dev-latest/tutorials/benchmarkApp_python/

 

Simply upload your model to the model/FP16 directory and update the following cell.

JesusE_Intel_0-1682710128773.png

 

Regards,

Jesus

 

0 Kudos
JesusE_Intel
Moderator
1,262 Views

If you need any additional information, please submit a new question as this thread will no longer be monitored.


0 Kudos
Reply