See an example of a DAG network used for a semantic segmentation application. Directed acyclic graph networks include popular networks, such as ResNet and GoogLeNet, for image classification or SegNet for semantic segmentation.
Using the cnncodegen function in GPU Coder™, you can generate CUDA® code from the semantic segmentation network. You can build it into a MEX function and generate code that runs 6x faster, at about 39 frames per second, than in MATLAB®. You can also integrate the generated code into a bigger application and deploy your deep learning algorithm on embedded GPU targets such as Jetson, Tegra®, or DRIVE™ PX platforms.