Deep Leaning Demonstration
Semantic Segmentation (or pixel classification) partitions the image into semantically meaningful parts,and labels each part.
TI Deep learning Library (TIDL) accelerates Semantic Segmentation implementation using variety of public/custom networks trained using open source frameworks.
1024x512 resolution Network with 15 fps on TDA2(*4*900MHz EVE)
Semantic Segmentation Network wih 41 Layer.Designed for 8 classes,trained for 5 classes.(pedestrain,vehicles,roads,road sign ,background)