Deep Learning in MATLAB: From Algorithm Design to Embedded Deployment on NVIDIA Jetson

Explore deep learning techniques and deployment strategies using MATLAB and NVIDIA technologies in this comprehensive 59-minute webinar. Dive into transfer learning workflows, GPU coding for deployment, and MATLAB to CUDA compilation. Discover how GPU Coder generates CUDA from both scalarized and vectorized MATLAB code, and learn about configurable memory management modes and shared memory synthesis for stencil operations. Examine vision algorithms and their GPU Coder speedup compared to CPU performance. Follow the journey from algorithm design to embedded deployment on NVIDIA Jetson, with practical examples including fog rectification and lane detection using modified AlexNet. Compare deep learning inference performance on Jetson TX1 and Titan Xp GPU. Gain insights into designing DNNs in MATLAB and deploying them with GPU Coder, equipping yourself with valuable skills for cutting-edge deep learning applications.

Intro
Example: Transfer Learning Workflow
Example: Transfer Learning in MATLAB
GPU Coder for Deployment
MATLAB to CUDA compilation
GPU Coder Generates CUDA from (scalarized) MATLAB
GPU Coder Generates CUDA from (vectorized) MATLAB
Memory Management modes are configurable
Shared Memory Synthesis for Stencil Operations
Vision Algorithms: GPU Coder Speedup over C code on CPU
Algorithm Design to Embedded Deployment on NVIDIA Jetson
Example: Fog Rectification Algorithm
Example: Lane Detection using Modified AlexNet on Jetson TX1
Deep Learning Inference: Alexnet on NVIDIA Jetson TX1
Deep Learning Inference: Alexnet on NVIDIA Titan Xp GPU
Design DNNS in MATLAB and Deploy with GPU Coder
Check out Deep Learning in MATLAB and GPU Coder