This project utilizes a YOLOv8 pretrained model from Ultralytics to perform filtered object detection on images, videos, or real-time webcam feeds. The filtered detector focuses on specific classes of objects from the COCO dataset. The included classes can be easily customized to suit your application.

• Python 3.x
• OpenCV
• Numpy
• Ultralytics YOLO

Install dependencies using:

pip install opencv-python numpy
pip install 'git+https://github.com/ultralytics/yolov5.git'

Create a custom filter_classes list in the main.py file to specify the classes you want to detect. You can refer to the COCO dataset for a complete list of classes.

Example:

# Create a custom filter_classes list to include the classes you want to detect.
# You can refer to the COCO dataset for a complete list of classes: https://cocodataset.org/#explore
# Example classes: 'person', 'car'
filter_classes = ['person', 'car']
# More examples can be added: 'bird', 'dog', 'cat', 'bicycle', ...

Or utilize the defined lists used with the test files in this repository

Example:

image_test_filters = ['car', 'truck']
video_test_filters = ['chair', 'couch', 'potted plant', 'dining table', 'tv']
realtime_test_filters = ['cow', 'person', 'bottle', 'backpack', 'spoon', 'knife']

Initialize the FilteredDetector with the specified filter classes in the main.py file.

Example:

# Initialize the FilteredDetector with the specified filter classes
detector = FilteredDetector(filter_classes)

Uncomment the desired method in the main function to detect objects over an image file, video file, or real-time webcam feed.

Example:

# Uncomment one of the following lines to choose the detection method
# detector.detect_over_image('test_files/img.png')
# detector.detect_over_video_file('test_files/cows.mp4')
# detector.detect_over_realtime_feed()

Run the main.py file to see the filtered object detection in action.

• This project uses a pretrained YOLOv8 model from Ultralytics, trained on the COCO dataset.

• Customize the filter_classes list to include the specific classes you want to detect.

• Feel free to explore and expand the functionality based on your project requirements.

Jacob Pitsenberger December 5, 2023

This software is licensed under the MIT License. By using this software, you agree to comply with the terms outlined in the license.