https://github.com/jakubschwenkbeck/object-recognition-prediction

Java AI application for advanced real-time object tracking using a custom Convolutional Neural Network (CNN) written from scratch in Java. The application captures video from a webcam, performs AI-powered object classification and tracking, and displays the feed in a GUI window using OpenCV.
https://github.com/jakubschwenkbeck/object-recognition-prediction

Last synced: about 1 month ago
JSON representation

Java AI application for advanced real-time object tracking using a custom Convolutional Neural Network (CNN) written from scratch in Java. The application captures video from a webcam, performs AI-powered object classification and tracking, and displays the feed in a GUI window using OpenCV.

• Host: GitHub
• URL: https://github.com/jakubschwenkbeck/object-recognition-prediction
• Owner: JakubSchwenkbeck
• License: mit
• Created: 2024-08-29T16:58:51.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-10-16T21:10:06.000Z (8 months ago)
• Last Synced: 2025-04-08T12:04:17.670Z (2 months ago)
• Language: Java
• Homepage:
• Size: 151 KB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Object Tracking in Java using OpenCV (Work in progress)

This project is a Java application that utilizes **OpenCV** to capture video from a webcam, perform object tracking and movement prediction, and display the video feed in real-time. It uses key components such as OpenCV library loading, camera initialization, frame capture, and image display using **HighGui**. Additionally, the application integrates a custom Convolutional Neural Network (**CNN**) written from **scratch** in Java to classify and track objects in the video feed.



    

    



## Features

- **OpenCV Integration**: Uses OpenCV for real-time video processing.

- **HighGui Display**: Renders video frames with object tracking on the screen.

- **Object Tracking**: Drawing bounding boxes around detected objects with classification.

- **Custom CNN Model**: Utilizes a self-written Convolutional Neural Network (CNN) implemented from scratch in Java for object classification.

### OpenCV for Java

**OpenCV** (Open Source Computer Vision Library) provides tools and libraries for real-time computer vision. This project leverages OpenCV for:

- **Video Capture**: Capturing live video from the webcam using the `VideoCapture` class.

- **Image Processing**: Performing real-time image processing tasks such as resizing, color space conversion, and feature detection.

- **Object Detection**: Applying custom object detection algorithms or pre-trained models to identify objects within video frames.

- **Image Display**: Using the `HighGui` class to create windows and display images and video streams in real-time.

- **Mouse and Keyboard Handling**: Responding to user input for interaction with the video feed.

### Own CNN:

The custom Convolutional Neural Network (CNN) implemented in Java is designed to classify and track objects from the video feed. Below are the key details and components of the CNN:

1. **Convolutional Layers**:

   - **Purpose**: Extract features from input images through convolution operations.

   - **Details**: The network includes multiple convolutional layers with varying filter sizes, strides, and padding to capture different levels of abstraction in the image.

   - **Example**: A layer with 32 filters of size 3x3, followed by a ReLU activation function.

2. **Pooling Layers**:

   - **Max Pooling**: Reduces spatial dimensions while retaining the most significant features. For example, 2x2 max pooling layers are used to downsample feature maps.

   - **Average Pooling**: Used in conjunction with max pooling to average feature values over local regions, enhancing the robustness of feature extraction.

3. **Fully Connected Layers**:

   - **Purpose**: Flatten the output from convolutional and pooling layers and make final predictions based on extracted features.

   - **Details**: Includes one or more fully connected layers that output the final class scores and bounding box coordinates.

- Training/Testing Data : [PASCAL VOC Data](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/devkit_doc.pdf)

    

## Requirements

- **Java Development Kit (JDK)**: Ensure you have JDK 8 or higher installed.

- **OpenCV Library**: Download and set up OpenCV for Java.

## Setup Instructions

1. **Download and Install OpenCV**:

    - Download OpenCV from the [official website](https://opencv.org/).

    - Follow the instructions to set up OpenCV for Java, including adding the `opencv-.jar` file to your project and configuring the native library path.

2. **Clone the Repository**:

    ```sh

    git clone https://github.com/JakubSchwenkbeck/Object-Recognition-Prediction

    cd Object-Recognition-Prediction

    ```

3. **Open the Project**:

    - Open the project in your preferred IDE (e.g., IntelliJ IDEA or Eclipse).

    - Add the OpenCV JAR file to your project's build path.

    - Set the VM options to include the native library path:

      ```plaintext

      -Djava.library.path=

      ```

4. **Run the Application**:

    - Execute the `Main` class.

    - The application will test each component and display the video feed with object tracking.

## Usage

- The application captures video from your default webcam and displays it in a window.

- A sample bounding box is drawn on the video feed for demonstration purposes.

- Press the 'q' key to exit the application.

## Troubleshooting

- **Camera Not Found**: Ensure your webcam is connected and try using a different camera index.

- **Library Load Error**: Verify that the OpenCV native libraries are correctly configured in your project.

## License

This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.