Model AI - Identification and reading of license plates
The client provides technology solutions for parking in Japan
Technology stacks
Image Processing, Deep Learning (YOLOX, Swin-transformer, ONNX, OCR)
Development Team
2 AI engineer + 1 PM + 1 Comtor
Duration
3 months
CHALLENGE
The customer's old technology license plate reading software is limited in its ability to read & process information due to its complete dependence on the original database. Inability to self-study & flexible handling.
HBLAB's Solutions
Combining Image Process with Deep learning Model, which helps to increase self-learning and processing capabilities. Thereby increasing the accuracy and flexibility when updating information, systems for the software.
RESULTS
Reduce the amount of Data that must be input to train the input system.
- System Accuracy: 99.97%
- Processing time: 2s/1 image (on low-configuration devices - fanless case)