Skip to content

.tflite file for tensor lite #405

Description

@anh-ntn-amira

Hello everyone 👋,

I am currently working on integrating U²-Net into an Android project using TensorFlow Lite. My main goal is to achieve real-time background removal, where the output is the main subject (foreground) with the background removed or transparent.

I have tried converting the PyTorch → ONNX → TensorFlow → TFLite pipeline, but I am facing several issues (quantization, conversion errors, and unexpected internal errors when running on-device).

👉 Therefore, I would like to kindly ask if anyone already has a working .tflite version of U²-Net (either full or U²-Netp) that:

Runs on Android with TFLite

Outputs the segmentation mask or directly returns the subject with background removed

This would be extremely helpful for testing and benchmarking performance on Android devices.

If possible, please share:

A direct .tflite file (Float32 / Float16 / Int8)

Thanks a lot in advance 🙏

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions