This paper focuses on detecting sports types from images, which can be used for action recognition. Detecting small objects like athletes and sports equipment is challenging due to their varying colors, appearances, and distance from the camera. CNN is the most common detection tool for sports, but it struggles with image accuracy due to various angles and light conditions. To detect sports using YOLO, the study suggests a model that uses Darknet-53. This model attained a of 97.25% on test data and 97.25% on train data. This study has the potential to identify and categorize different sports and their analytical data. We were 97.25% correct in identifying the recommended model for altering the YOLO network for sports classification using game items. Broadcasting agencies are the most important prerequisite for preserving sports’ allure. Sports images can now be more easily recognized according to their different classes thanks to our research. There hasn’t been much research done, particularly in terms of identifying sports from the previously provided data. Not only were there few jobs in the sports industry overall, but no one worked in it. They concentrated on handball and basketball. Algorithm research in this area has the potential to enhance sports analytics and spectator services. The best pre-trained CNN model and classifier to apply in pertinent future studies projects are what we hope to achieve by running these experiments. Our research is probably going to lead to the development of better sports identification algorithms.