Detecting Collectable Resources on Computer Game Based on the Neural Network
Bibliography
Shen, Q. (2023). Detecting Collectable Resources on Computer Game Based on the Neural Network. Highlights in Science, Engineering and Technology, 34, 226–231. https://doi.org/10.54097/hset.v34i.5476
Abstract
Image recognition and classification were put in used in many places in recent years thanks to the great progress in machine learning and neural networks. However, seldom evidence indicate applications of image recognition in computer games. Since games are gradually becoming part of peoples life, it is the time to put the focus on those who are unable to enjoy this entertainment. Nowadays computer games often include the element of collecting e.g. (materials or treasure). A program that could automatically mark those collectable objects would significantly improve the gaming experience for those who have disability in their eyes. In this work the game Destiny 2 was used for training and testing and training model chose was Yolov5. 1,500 of images were put into the data set and trained with 350 epochs. The model eventually achieved 0.82 in precision, 0.96 in recall and 0.91 in F1. It also gets the average mAP value of 0.93. These results can strongly prove that this model has the potential to be used in real scenarios.