Football Team Foul Detection

Intro to Deep Learning Project CS492, Fall 2021, KAIST.

Game Dataset

To install dataset, run Dataset/download_Soccernet.py and auto_trim.py

In each game folder, / Dataset/SoccerNet / trim_dir / league / year / match

• '1_field_calib_ccbv.json',
• '1_HQ.mkv', => First Half, Full Match
• '1_player_boundingbox_maskrcnn.json', => RCNN Prediction from Giancos's data, 1st Half
• '1_ResNET_TF2.npy',
• '1_ResNET_TF2_PCA512.npy', => After PCA
• '2_field_calib_ccbv.json',
• '2_HQ.mkv',
• '2_player_boundingbox_maskrcnn.json',
• '2_ResNET_TF2.npy',
• '2_ResNET_TF2_PCA512.npy',
• 'annotator.txt',
• 'Foul', => Trimmed Videos of Foul
• 'Labels-cameras.json', => Containing camera view(Close up player/main referee, Main Camera), time
• 'Labels-captioning.json', => Match Commentary, inaccurate
• 'Labels-v2.json', => Containing all events(kick-off, foul, card, etc) and which team commiting the events
• 'Labels.json', => Containing Events (Kickoff, Card, End of Match, Substitution) only
• 'labels_event.json', => Containing Foul, Yellow and Red card and Which team commits the foul
• 'Red card', => Trimmed Videos for Red Card
• 'video.ini', => Details about 1_HQ.mkv and 2_HQ.mkv
• 'video_with_duration.ini', => Details about 1_HQ.mkv and 2_HQ.mkv
• 'Yellow card' => Trimmed Videos of Yellow Card