Using the Keras deep learning framework, I was able to train a convolutional neural network (CNN) to classify images into 100 different sports. Originally I tried to train two models from scratch, but fialed to reach an adequate validation score due to overfitting, lack of data, etc. Instead, I chose to pursue the transfer learning route and trained a model on the ResNetMobileV2 CNN, with data augmentation, and was able to achieve a test-accuracy score of approximately 78%.
https://www.kaggle.com/datasets/gpiosenka/sports-classification?datasetId=1209061&sortBy=voteCount
• Python (Keras, Pandas, Matplotlib)