Autodetectify

Welcome to Autodetectify, an advanced solution for for testing Zero-Shot(ZS) object detection models with different prompts and autolabelling datasets. This application provides a modern interface for comparing different object detection prompts on your uploaded images, auto-labeling images based on your specified prompt, and exporting the dataset in various formats.

Features

• Prompt Comparison: Upload your images, enter different prompts, and visualize side-by-side comparisons of object detection results.
• Auto-Labeling and Export: Automatically label your images using a specified prompt and download the labeled dataset in YOLOv5, YOLOv8, or COCO format.

How to Use

Prompt Comparison

1. Navigate to the Prompt Comparison tab.
2. Upload your images.
3. Enter the prompts you want to compare.
4. View and compare the object detection results side by side.

Auto-Label and Export

1. Navigate to the Auto-Label and Export tab.
2. Upload up to 50 images or a ZIP file containing images.
3. Enter the prompt for labeling.
4. Run the auto-labeling process.
5. Download the labeled dataset in the desired format.

Installation

To run this application locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/sefaburakokcu/autodetectify
   cd autodetectify

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the Streamlit app under ui:

   streamlit run 🤗_Home.py

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

• Streamlit
• streamlit-image-comparison
• transformers