Researchers have developed an embedded real-time license plate recognition system tailored for developing countries, addressing complex traffic scenes and diverse vehicle types. The system utilizes lightweight convolutional neural networks for both license plate detection and character recognition, achieving 93.6% mAP and 87.88% accuracy on the new SL-LPR dataset. To ensure efficiency on embedded platforms, the models incorporate low-bitwidth quantization via Brevitas and FPGA acceleration through the FINN framework, enabling operation at 11.5 FPS on a Xilinx Kria KV260. AI
IMPACT This research demonstrates efficient AI model deployment on embedded systems for real-world applications like traffic management.
RANK_REASON The cluster contains a research paper detailing a new system and dataset.
- brevitas
- Finn
- SL-LPR dataset
- Sri Lanka
- Udaya Sampath Karunathilaka Perera Miriya Thanthrige
- Xilinx Kria KV260
- SL-LPR
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
