Updates from Coral: A new compiler and much more

Vikram Tank (Product Manager), Coral Team
3 min readintermediate
--
View Original

Overview

The article discusses recent updates from Coral, including the launch of a new Edge TPU model compiler, enhancements to Mendel OS, and the introduction of new hardware like the Environmental Sensor Board. These updates aim to improve the developer experience and expand the capabilities of on-device AI applications.

What You'll Learn

1

How to utilize the new Edge TPU model compiler for various architectures

2

Why the Mendel Development Tool (MDT) enhances board management for developers

3

How to run multiple models in parallel on Edge TPU devices using C++ and Python APIs

4

When to leverage the Environmental Sensor Board for enhanced sensor input in AI models

Key Questions Answered

What improvements have been made to the Edge TPU model compiler?
The Edge TPU model compiler has been updated to remove restrictions on specific architectures, allowing developers to submit any model architecture. This change significantly increases the variety of models that can be run on the Coral platform.
What new features does the Mendel OS 3.0 Chef include?
Mendel OS 3.0 Chef introduces a new board management tool called the Mendel Development Tool (MDT), which streamlines the management of the Coral Dev Board, enhancing the developer workflow.
How can developers run multiple models in parallel on Edge TPU?
Developers can now use both the C++ and Python APIs to run multiple models in parallel on multiple Edge TPU devices, improving the efficiency of AI applications.
What capabilities does the Environmental Sensor Board provide?
The Environmental Sensor Board is an accessory for the Coral Dev Platform that includes integrated sensors for light, temperature, humidity, and barometric pressure, along with additional Grove connectors for more sensors.

Technologies & Tools

Hardware
Edge Tpu
Used for running AI models efficiently on-device.
Operating System
Mendel OS
Provides the software environment for Coral devices.
Framework
Tensorflow Lite
Facilitates running machine learning models on Edge TPU.

Key Actionable Insights

1
Leverage the new Edge TPU model compiler to explore a wider range of model architectures for your AI applications.
This update allows developers to experiment with various architectures that were previously restricted, thus enhancing the flexibility and capability of their models.
2
Utilize the Mendel Development Tool (MDT) to streamline your development process on the Coral Dev Board.
The MDT simplifies board management, making it easier for developers to deploy and manage their applications effectively.
3
Incorporate the Environmental Sensor Board into your projects to gather real-time environmental data.
This board enhances the Coral platform's capabilities by providing essential sensor data, which can be crucial for developing responsive AI applications.

Related Concepts

AI
Iot
Tensorflow Lite
Coral
Machine Learning