Yolov4 artificial neural network, trained to detect and count champions health bars on the League of Legends game image files.
The artificial neural network was trained with help of the Darknet framework, using Google Colaboratory.
In order to collect dataset, jpg images were extracted from gameplay videos with a frequency of 0.2 fps. Over 2100 images were collected.
Images extraction was realized, using extract_frames.py
script.
Labels for the images were created manually, using LabelImg program. Example dataset is presented in the example_dataset folder.
To plot every image from given directory with bounding boxes and total score, run run_yolo_detection.py
script and enter images directory.
To print total number of champions health bars in the image, run run_yolo_count.py
.
Pass images names as a command line arguments e.g. python run_yolo_count.py img1.jpg img2.jpg img3.jpg