EITNet: An IoT-Enhanced Framework for Real-Time Basketball Action Recognition

TL;DR

EITNet tackles real-time basketball action recognition under occlusion and high-speed motion by fusing multi-view video data with IoT sensors. The approach combines EfficientDet for fast player detection, I3D for spatiotemporal feature extraction, and TimeSformer for temporal analysis, all enabled by real-time IoT data collection. Key contributions include demonstrating a 0.92 accuracy with loss below 5.0 over 50 epochs, and showing strong cross-subject and cross-view generalization with low model complexity. The work advances automated sports analytics by enabling real-time feedback and data-driven training optimization in practical basketball settings.

Abstract

Integrating IoT technology into basketball action recognition enhances sports analytics, providing crucial insights into player performance and game strategy. However, existing methods often fall short in terms of accuracy and efficiency, particularly in complex, real-time environments where player movements are frequently occluded or involve intricate interactions. To overcome these challenges, we propose the EITNet model, a deep learning framework that combines EfficientDet for object detection, I3D for spatiotemporal feature extraction, and TimeSformer for temporal analysis, all integrated with IoT technology for seamless real-time data collection and processing. Our contributions include developing a robust architecture that improves recognition accuracy to 92\%, surpassing the baseline EfficientDet model's 87\%, and reducing loss to below 5.0 compared to EfficientDet's 9.0 over 50 epochs. Furthermore, the integration of IoT technology enhances real-time data processing, providing adaptive insights into player performance and strategy. The paper details the design and implementation of EITNet, experimental validation, and a comprehensive evaluation against existing models. The results demonstrate EITNet's potential to significantly advance automated sports analysis and optimize data utilization for player performance and strategy improvement.

Figures (7)

• Figure 1: Overall architecture of the EITNet model, integrating EfficientDet for object detection, Inflated 3D CNN for spatiotemporal feature extraction, and TimeSformer for temporal sequence analysis in action detection and classification. The model leverages IoT integration for real-time data acquisition and feedback, enabling accurate recognition of complex basketball actions across multiple views.
• Figure 2: Flowchart of the integrated IoT system for basketball player action recognition, showing the process from multi-camera data collection to teaching design, real-time action recognition, and prediction, with data managed by an intelligent computer and teaching system.
• Figure 3: Architecture of EfficientNet, showing the progression from the input layer through various feature pyramid levels (P2 to P7) into the BiFAN Layer for enhanced feature fusion, followed by the head for final convolutional processing.
• Figure 4: Structure of the Inflated 3D Convolutional Network, illustrating the Inflated Inception-V1 architecture with detailed Inception modules for spatiotemporal feature extraction.
• Figure 5: Architecture of the TimeSformer model, showing the process from input embedding through divided space-time attention, linear transformation, and softmax output probabilities.