An Embedded AI Design Methodology and Use Cases for ADAS Applications
![Printer-friendly version Printer-friendly version](/sites/all/modules/print/icons/print_icon.gif)
![Send by email Send by email](/sites/all/modules/print/icons/mail_icon.gif)
![PDF version PDF version](/sites/all/modules/print/icons/pdf_icon.gif)
Prof. Jiun-In Guo, Distinguished Professor
Institute of Electronics, National Yang Ming Chiao Tung Universit, TW
Abstract
This talk discusses our in-house embedded AI design methodology and some usecases for ADAS applications. The proposed methodology comprises an automatic model pruning method to simplify the model complexity according to user’s demand in terms of different pruning ratios and then followed by an automatic self-learning method to fine tune the simplified model to be adapted to field applications without labeling datasets. Then, an automatic model quantization method is applied to refine the fine-tuned model from 32-b float weights to 8-b/4-b weights before being deployed on an embedded AI SoC with limited quality degradation. Finally, some ADAS usecases applying the proposed embedded AI design methodology are introduced to show its model compression performance.
Contact information
http://ivs.ee.nctu.edu.tw/ivs/index.php/advisor
Email: jiguoccugmail [dot] com