Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Strange behaviour with OpenVINO and int64 custom layers #25533

Open
4 tasks done
alexlyulkov opened this issue May 3, 2024 · 0 comments
Open
4 tasks done

Strange behaviour with OpenVINO and int64 custom layers #25533

alexlyulkov opened this issue May 3, 2024 · 0 comments
Labels

Comments

@alexlyulkov
Copy link
Contributor

System Information

OpenCV 5.x
Ubuntu 22.04
OpenVINO 2024.0.0

Detailed description

Sometimes OpenVINO pipeline in OpenCV changes the type of the model's output tensor from int64 to int32.

OpenVINO supports int64 tensors, but it doesn't support int64 operations, so it converts all int64 internal tensors to int32. So the models with int64 inputs or outputs work fine, but with int32 real precision. However, when we use custom OpenCV layers in OpenVINO pipeline, the OpenVINO compiler changes the type of the model's output tensor from int64 to int32.

Steps to reproduce

Enable Test_ArgMax_Int.random test for OpenVINO backend

Issue submission checklist

  • I report the issue, it's not a question
  • I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
  • I updated to the latest OpenCV version and the issue is still there
  • There is reproducer code and related data files (videos, images, onnx, etc)
@alexlyulkov alexlyulkov added the bug label May 3, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant