ML Examples
ML Examples is a collection of source codes for machine learning tutorials and examples that are part of Arm's ML developer space. This project is aimed at providing developers with practical tools and guides for implementing machine learning solutions on various Arm platforms. Here is an overview of some available projects and tutorials:
Arm NN Mobilenet on Android
This project focuses on deploying a quantized TensorFlowLite MobileNet V2 model on Android devices using the Arm NN SDK. It offers a straightforward way to integrate efficient image classification capabilities into mobile applications. The source code for this tutorial can be found on GitHub.
Arm Style Transfer on Android
This project implements neural style transfer on Android devices using Arm NN APIs. The tutorial guides developers through the process of applying artistic styles to images, a technique that can be used in photography apps or creative tools. More details are available in the tutorial and the source code is hosted on GitHub.
CMSIS Pack based examples for Arm Corstone 300
This project showcases keyword spotting and object detection capabilities on the Arm® Corstone™-300 platform. It highlights the use of CMSIS projects to create efficient embedded ML applications. Developers can access the source code on GitHub.
Ethos-U on Corstone 300
Focusing on exploring the Arm® Corstone™-300, this project utilizes the Arm® Cortex™-M55 and Arm® Ethos™-U55 NPU for efficient machine learning inference. It is ideal for developers interested in leveraging cutting-edge NPU technology. The project's source code is also available on GitHub.
Multi-Gesture Recognition
The multi-gesture recognition tutorial guides developers through training a convolutional neural network to recognize gestures under a variety of conditions using TensorFlow and a Raspberry Pi. It provides extensive insights into implementing gesture control solutions. More information is available in the tutorial and its GitHub repository.
Fire Detection on a Raspberry Pi using PyArmNN
This project involves deploying a neural network to detect fire or flames in images on a Raspberry Pi device using PyArmNN. It emphasizes rapid machine learning inference and can be particularly useful for safety and surveillance applications. The tutorial and source code on GitHub offer further details.
PyTorch to Tensorflow
This project provides a Jupyter notebook that demonstrates how to convert models trained in PyTorch to TensorFlow Lite format. This is valuable for developers looking to transition between these popular machine learning frameworks, providing flexibility in deploying models. The source code can be found on GitHub.
RNN Unrolling for Tf Lite
This tutorial offers insights into training Recurrent Neural Networks (RNNs) with TensorFlow and preparing them for deployment in TensorFlow Lite format. It is particularly interesting for those working with sequential data and requiring compact model deployment. The source code is accessible on GitHub.
Image Recognition on MBED using CMSIS and TFLM
This project provides an image recognition demo on a Discovery STM32F746G board utilizing TensorFlow Lite for Microcontrollers (TFLM) and CMSIS-NN. It demonstrates deploying efficient image recognition in microcontroller environments. The tutorial can be viewed here along with the source code on GitHub.
Yeah, World
This project involves gesture recognition using TensorFlow and transfer learning on the Raspberry Pi platform, including the Pi 4 Model B, Pi 3, and Pi Zero. The tutorial and source code provide a means of exploring machine learning capabilities in IoT devices, accessible on GitHub.
These projects collectively enable developers to harness the power of machine learning in various applications, demonstrating the adaptability and utility of Arm's technologies in a growing array of devices.