Real-time classification of handball game situations

IEEE ICTAI 2022 — 34th International Conference on Tools with Artificial Intelligence

Presentation of a real-time deep learning system achieving 98.6% classification accuracy for handball broadcast automation, published and presented at IEEE ICTAI 2022 in Macao.

Presented at the 34th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2022) in Macao, China.

This paper introduced a deep learning classifier for real-time handball game situation recognition, enabling automated broadcast decisions without human intervention. The model processes over 25 frames per second with only 34 ms end-to-end latency — well within live production constraints.

Key contributions presented:

A CNN-based classifier trained on 127,600 labeled handball frames
Integration into a real-time pipeline meeting live broadcast timing requirements
98.6% accuracy on unseen matches from the same team
Demonstration of feasibility for AI-assisted autonomous match direction