Context This dataset was created to train neural networks to categorize juggling patterns based on the positions of the balls and the juggler's hands. The accompanying code can be found on github: https://github.com/rasmusakerlund/juggling-vision-py Content The dataset consists of 36 csv-files with 3000 rows each. Each row represents one frame at 30 frames per second and records the x and y coordinates of the right hand followed by the left hand and then followed by the balls. The dataset is presented visually in this video: https://youtu.be/LLeCsy3BSHs