CV Toolkit

Tuning OpenCV pipelines is usually a huge chore, involving enabling/disabling transformations, tuning values, redeploying, and tons of boilerplate code. We sought to make this process easier with our open-source CV Toolkit. Users can:

• Upload a source image

• Toggle processing steps on/off

• Modify step settings through UI controls

• Save current pipeline + parameters to JSON

• Reload configurations later to continue from where they left off

• Steps are Python files representing individual CV operations

• JSON files represent pipelines and store:

  • • step metadata (name, enabled state, etc.)

    • parameter values (e.g., kernel size, blur intensity)

    • setting input types (slider, text input, and more)

• The CV Toolkit comes with over 10 different steps, but you can always add more. We highly recommend using agentic AI to create steps based off of our examples.

1. Clone the `cv-toolkit` repository from our GitHub and open it using the command line or an IDE (like VSCode)

2. Create a virtual environment using `python -m venv .venv` (Windows) or `python3 -m venv .venv` (Mac/Linux)

3. Launch the virtual environment with `source .venv/bin/activate`

4. Install dependencies using `pip install -r requirements.txt`

5. Launch the program using `streamlit run main.py`

6. Load a pipeline JSON file. You can start with a provided example JSON or create one from a template. Use the `Load` button to upload the file.

7. Tune your pipeline! Enable/disable steps, adjust settings, preview the output, and export your JSON once you're done. You can always load it later if you need to.