Baseball Game Event Detection Model
Summary I built a pipeline for automated event detection in baseball broadcasts. The model combines a 3D CNN backbone with a motion branch using temporal convolutions to better capture player and ball dynamics over time.
Highlights
3D spatiotemporal features for broadcast clips
Motion branch improves detection of transient actions
Train-eval scripts with clear dataset interface
Stack Python, PyTorch or TensorFlow style workflows, NumPy, ffmpeg
Links
Repo: https://github.com/Andrew-Bumgardner/cmsc472-final_proj/
Project link: https://github.com/Andrew-Bumgardner/cmsc472-final_proj/