This project implements object detection using YOLO (You Only Look Once) in Python, with a simple GUI interface to choose between game/screen or camera-based detection. The system uses YOLOv3 for object detection and allows users to detect objects such as cars, people, and more in real-time from games, screens, or a connected camera.

• Real-time object detection using YOLOv3.
• GUI interface to select detection mode: either from a live webcam feed or game/screen capture.
• Supports various COCO dataset objects such as cars, people, buses, and more.
• Click-based interaction when detecting objects on the screen.

To get started with this project, ensure you have the following dependencies installed:

pip install -r requirements.txt

• Python 3.x
• OpenCV (opencv-python)
• NumPy
• Tkinter (for the GUI)
• PIL (Pillow)
• PyAutoGUI (for screen capture)
• Pynput (for mouse control)

1. Clone this repository:

git clone https://github.com/yourusername/yolo-game-screen-detection.git
cd yolo-game-screen-detection

2. Install the required Python packages:

pip install -r requirements.txt

3. Download the YOLOv3 weights and configuration files.

You can download the pre-trained YOLOv3 weights using the links below:

• YOLOv3 Weights: Download YOLOv3 weights
• YOLOv3 Configuration: Download YOLOv3 config

Once downloaded, place the weights and configuration files in the project folder or adjust the paths in the code if stored elsewhere.

1. Launch the GUI to select the detection mode (Game/Screen or Camera):

python main.py

2. In the GUI:

   • Choose between Game/Screen or Camera for the source of object detection.
   • Press Start Detection to begin real-time object detection.
   • For Game/Screen, the system will capture the current screen for detection.
   • For Camera, it will use your webcam feed.

3. Press the f key during detection to toggle click-based interaction.

This project utilizes OpenCV, YOLOv3, and PyAutoGUI for screen and video capture. Below is an example of how the YOLO object detection process is handled in the code:

# YOLO model files
net = cv2.dnn.readNet("yolov3.weights", "yolov3.cfg")

# Get YOLO output layer names
layer_names = net.getLayerNames()
output_layers = [layer_names[i - 1] for i in net.getUnconnectedOutLayers()]

# Define object classes (COCO dataset)
classes = ["person", "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", ...]

• To adjust the objects detected, you can modify the classes list in the code to include only the objects you are interested in detecting.
• Adjust the detection confidence threshold in the code to filter out low-confidence detections:

if confidence > 0.5:
    # Process detected object

This project uses the YOLO implementation by Joseph Redmon and integrates various libraries such as OpenCV, PyAutoGUI, and Tkinter.

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