Real-time Object Detection with Object-Orion

This project implements a real-time object detection application using a custom YOLO-based model called Object-Orion. The application is built with Streamlit and offers multiple input options including image upload, video upload, and real-time camera feed.

Try out the app. object-orion

1. Camera module doesnt work on web yet. You can try running it on your system.
2. Download the code.
3. Open the Object-Detection-Using-YOLO directory in any terminal.
4. Run the commands: pip install -r requirements.txt streamlit run app.py

Voila!

Features

• Real-time object detection using Object-Orion v1
• Support for image, video, and camera input
• User-friendly interface built with Streamlit
• Asynchronous processing for smooth performance

Dataset : Pascal VOC

The Pascal VOC dataset is a benchmark in computer vision, containing annotated images for object detection and classification tasks, with defined 20 object categories. Pascal-VOC

Model: Object-Orion

Object-Orion is a custom object detection model based on the YOLO (You Only Look Once) architecture. It's currently in its first version (v1) with an accuracy of approximately 69%. While it's still in development, it demonstrates promising results for real-time object detection tasks.

For more information about Object-Orion, visit Object-Orion.

Project Structure

The project consists of two main Python files:

1. app.py: The main application file that creates the Streamlit interface and handles user interactions.
2. yolo_predictions.py: Contains the YOLO_Pred class that interfaces with the Object-Orion model for making predictions.

app.py

This file sets up the Streamlit application and includes the following key components:

• YOLOTransformer class: Handles real-time video processing using the Object-Orion model.
• Input selection: Allows users to choose between image, video, or camera input.
• Image and video processing functions: Handle uploaded images and videos.
• WebRTC implementation: Enables real-time camera feed processing.

yolo_predictions.py

This file contains the YOLO_Pred class, which is responsible for:

• Loading the Object-Orion model
• Processing input images or video frames
• Applying non-maximum suppression to filter detections
• Drawing bounding boxes and labels on the processed images

Usage

1. Install the required dependencies:

2. Run the Streamlit app:

3. Select your preferred input method (Image, Video, or Camera) from the radio buttons.

4. For image or video input, upload a file using the file uploader.

5. For camera input, allow the application to access your camera when prompted.

6. View the results of object detection in real-time!

Future Improvements

• Improve the accuracy of the Object-Orion model
• Add support for more object classes
• Optimize performance for faster processing
• Implement object tracking for video and camera inputs

Contributing

Contributions to improve the application or the Object-Orion model are welcome. Please feel free to submit pull requests or open issues on the GitHub repository.

License

MIT LICENSE