Skip to content

Detect various kind of data of badminton matches using Bidirectional LSTM || AI Cup 2023 - Teaching Computers to Watch Badminton Matches (5th place) / 教電腦看羽球 (第五名)

Notifications You must be signed in to change notification settings

aisu-programming/CV-Badminton-Matches

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CV-Badminton-Matches

AI CUP 2023 - Teaching Computers to Watch Badminton Matches

Awards: https://global.turingcerts.com/en/co/cert?hash=0x542087a0cba32a7e5cd03330a9bb02c46ba0fe8a437f0ec52ecf60b1274260780000

Model Architecture

image

Install

  1. Run install_pytorch.sh or install the version you like
  2. Run install_mm.sh or manually install MMDetection and MMPose
  3. Run install_tracknetv2.sh or manually install TrackNetv2

Training Procedure

  1. Background Extraction (1_background_extraction.py)
  2. Background Clustering (2_background_clustering.py)
  3. Ball Detection (3_ball_processing.py)
  4. Player Pose Detection (4_pose_detection_*.py)
  5. Train models for each columns (5_train_*.py)

Predicting Procedure

Predict answers of each columns by each models (6_predict_*.py) step by step

Ideas

Court Detection

(turned out to be unnecessary)

  1. Edge detection (Canny): X
  2. Homography: X
  3. DIY: O

Background Extraction

  1. DIY - "average_method": X
  2. DIY - "mode_method": X
  3. DIY - "average_method_with_masked_players": O

Ball Detection

  1. Blob detection: X
  2. TrackNetv2: X
  3. TrackNetv2 + postprocess: O

Player Pose Detection

(Backgound Extraction finished needed)

Others

More content in my report but in Chinese version

About

Detect various kind of data of badminton matches using Bidirectional LSTM || AI Cup 2023 - Teaching Computers to Watch Badminton Matches (5th place) / 教電腦看羽球 (第五名)

Topics

Resources

Stars

Watchers

Forks