Metadata-Version: 2.1
Name: arm_segmentation
Version: 0.0.1
Summary: Training and inference of instance segmentation
Author-email: Peter Mitrano <pmitrano@umich.edu>
Maintainer-email: Peter Mitrano <pmitrano@umich.edu>
Requires-Python: >=3.6
Description-Content-Type: text/markdown
Provides-Extra: test

[![Python package](https://github.com/UM-ARM-Lab/arm_segmentation/actions/workflows/python-package.yml/badge.svg)](https://github.com/UM-ARM-Lab/arm_segmentation/actions/workflows/python-package.yml)
[![Upload Python Package](https://github.com/UM-ARM-Lab/arm_segmentation/actions/workflows/python-publish.yml/badge.svg)](https://github.com/UM-ARM-Lab/arm_segmentation/actions/workflows/python-publish.yml)

This repository contains very simple training, inference, and visualization code for instance segmentation.
This is a good choice for when you don't want to get into the details and just want a model that works well and is cheap
and fast (compared to services like AWS)

# Data

This repository assumes you have used RoboFlow to annotate your dataset.
You should export your dataset as a COCO style dataset, which stores annotations in JSON.

# Installation & Setup

1. Clone this repository
   ```
   git clone git@github.com:UM-ARM-Lab/arm_segmentation.git
   ```
2. In an existing or new python virtual environment, install the dependencies
    ```
    pip install -r requirements.txt
    ```
3. Train
    ```
    # You may need export PYTHONPATH=./src, or you can try pip install -e .
    ./scripts/train.py path_to_dataset
    ./scripts/visualize_validation.py path_to_dataset
   ```
