python-object-detection-with-yolo-and-opencv

Table of Contents

• Summary
• References
• How to use

Summary

Simple Object detector app utilizing trained YOLO v3 or YOLO v4 CNN models.

References

• darknet Github repository
• yolov4_darknet Github repository
• YOLO: Real Time Object detection - The tutorial also explains how supervised learning can be applied for the YOLO architecture to generate a custom trained model with .weights (weights of the nodes determined by backward pass), .cfg (contains network architecture details, hyperparameters, input and output configuration, training settings) and .names (model classes utilized for training) files.

How to use

Perequisite

• Ensure you have a WebCam

Install the pip packages

pip install -r requirements.txt

Retrieve files required for object detection

Following links can be utilized to download the .weights files:

• yolov3.weights
• yolov3-tiny.weights
• yolov4.weights
• yolov4-tiny.weights

Copy those weights in the cfg folder in this project.

(Optional) Copy .cfg and .names files

Following steps can be considered in order to retrieve the.cfg and .names files:

• Clone both mentioned Github repositories in the Reference section
• In darknet/cfg and yolov4_darknet/cfg you can find the cfg files which you can copy to the cfg folder if not already existing
• In darknet/data and yolov4_darknet/data you can find the .names (list of classes) files, e.g. coco.names

Run the app

Execute the object detection application utilizing any of the downloaded YOLO weights:

python object_detector_app.py --help
# Object detection with yolov4
python object_detector_app.py --model yolov4-tiny