Guide to Porting Real-time Traffic Sign Recognition Algorithm on Texas Instruments TDA2x SoC
Autonomous systems when plying on the streets need precise information pertaining to their surroundings and traffic signs that are on the road. The go to solution for identifying traffic sign recognition is advanced automotive camera coupled with CNN models that are trained with diverse dataset. These algorithms are computation intensive and are difficult to have a working demonstration on automotive embedded platform. This article explains in detail steps to be followed in porting these advanced CNN models on Texas Instruments TDA2x SoC. This article sheds light on following aspects involved in porting.
- Various challenges involved in porting of these algorithms like- Model conversion, Model customization, Model complexity and TIDL inference adaptions
- Demonstrating TSR on TDA2x