Machine Learning using Arm
Delivery method: Online video
Course Length: 1 hour
Technology Focus: Combined Hardware and Software
This training topic covers essential information on Arm’s IP solutions for optimizing Machine Learning (ML) applications for Arm hardware. The topic introduces Arm’s solutions for implementing ML on Android and Linux platforms with our suite of ML software tools and how they relate to corresponding Arm hardware.
The topic provides an overview of Arm Compute Library, an open source toolkit for implementing ML applications on Arm’s Mali GPU’s.
Because Arm’s ML solution is optimized for Convolutional Neural Networks (CNNs), the course also provides an introduction to (CNNs) and how they operate. Finally, the course has some tutorials for you to follow that demonstrate how to run some ML example applications on Arm hardware, including Arm Cortex-A, Arm Cortex-M, and Arm Mali GPUs.
Here is a short preview of the course:
- A basic understanding of neural networks.
- The course is relevant to anyone who wants to start targeting Machine Learning applications on Arm hardware IP.
- 1 hour
Arm IP and Machine Learning
- Intro to Arm's machine learning IP
- Arm SW Library - Arm NN
- Arm SW Library - ACL and CMSIS-NN
- Arm ML Hardware
Introduction to Machine Learning
- Introducing Convolutional Neural Networks (CNNs)
- How does convolution work?
- Convolution - stride and padding
- Convolution - channels
- Convolutional Neural Networks (CNNS) in action
- Transfer learning
Introduction to Object Detection
- Introduction to Object Detection
- Bounding boxes
- Sliding windows
- Region proposals
- Landmark detection
- YOLO - You Only Look Once
- Anchor boxes
- Intro to OpenCL
- OpenCL Execution Model
- OpenCL example
How-to run ML tasks on Arm hardware
- Image recognition on Arm Cortex-M with CMSIS-NN
- Running and profiling Arm NN on the HiKey 960 (Arm Cortex-A)
- Running and profiling Arm NN on the HiKey 960 (Arm Mali GPU)