Next-Generation Parallel Decoder for LPDR: Architectural Optimization and Class-Balanced GAN-Augmentation
Researchers have developed a new method to improve real-time license plate detection and recognition (LPDR) systems. The approach addresses issues of spatial character mismatches and data imbalance in training sets by introducing Cross-Spatial Hybrid Attention (CSHA) and Class-Balanced Synthetic Augmentation (CBSA). This technique significantly boosts the recognition rate for less common license plates, from 78.2% to 91.5%, while maintaining high processing speeds of 152 FPS. AI
IMPACT Enhances real-time AI applications in smart cities by improving accuracy and efficiency for license plate recognition systems.