Instructions for use

Overview

This repository provides functionality for a 3 step conversion process to convert the U2Net model from its original PyTorch format to TensorFlowLite format.

Pre-Requisites

1. Install the required dependencies via pip install -r requirements.txt
2. Download the pre-trained u2net model here and place it in content/.
3. Place an image of choice(.jpg) in content/
4. Run the pth_onnx script. The output model should appear in the root directory.
5. Verify the successfulness of the conversion by examining the console statements.
6. Run the convert_onnx_tf script. The converted model will appear in the root directory.
7. Run the convert_tf_tflite script. The converted model will appear in the root directory.

Notes

The repository also includes model-dimensions.py, which is a python script that prints the input requirements of a model. This is to facilitiate ease of integration of the model into your program.

Credits

Source code taken from farmaker47, which can be found here