Our aim for Roboflow Evaluations is to create a one-stop tool for evaluating computer vision model performance, with a specific emphasis on comparing model predictions to ground truth.
We welcome contributions to the library to help expand support for new models and to build new evaluation tools useful for engineers.
There are a few ways to contribute to Roboflow Evaluations:
- Add support for a new model.
- Add a new data loader for taking data from a generic format and ingesting it into the standard data loader format.
Before you submit a pull request (PR) that adds a feature to the library, please create an Issue with a feature request.
This gives all contributors an opportunity to weigh in on your idea. This process may help facilitate more ideas. Further, submitting a feature request before any code changes are made increases the chance that your idea will be implemented in the library.
Whether you want to improve documentation, submit a bug fix, or add to the library, you'll first need to get our repo set up on your local machine.
First, fork the repository to your own GitHub account. This fork is where you will push your code changes.
Next, you'll need to download and configure the project on your local machine. You can do this by following the instructions in our README file. The README outlines how to install the library and run test cases.
Please create a new branch for your changes by using the git checkout -b <branch-name> command.
Once you have made a change, please run all test cases according to the instructions in the README file. This helps us assure the quality of the code submitted to the library.
When you are ready to submit a change, commit your changes to the branch. Then, submit a pull request to this repository.
A contributor will review your request and either approve it or provide feedback and proposed changes.
Do you have some feedback on how we can improve this file? Let us know by submitting an issue in the Issues section of this repository.