Together with Digi-Key, Hackaday is proud to present the Train All the Things Contest. The goal is to come up with a project that uses Machine Learning in the weirdest, wildest way possible. As a subset of Artificial Intelligence, Machine Learning gives you access to do things that could never be accomplished before. Here's your chance to turn that power into something people will use every day - or never use at all! - and challenge yourself and your machine.

Thank you to everyone who entered! The winners for each category are the following:

Machine Learning on the Edge -- Intelligent Bat Detector

Machine Learning on the Gateway -- AI Powered Bull**** Detector

Artificial Intelligence Blinky -- Soldering Lightsaber

Machine Learning in the Cloud -- Hacking Wearables for Mental Health and More

How is this different than Artificial Intelligence? Artificial Intelligence encompasses Machine Learning in that it refers to all kinds of ways that machines can emulate human processing.

For the Train All The Things contest, we’re looking for the weird, the useful, the thing you'll use every day, and the thing you'll never use. Most of all, we’re looking for you to challenge yourself and your machine. 

You can write machine learning algorithms on many types of hardware: Beaglebone, NVdia Jetson Nano, Raspberry Pi, STM32, and Arduino. Machine Learning can encompass, among others, image classification, object detection, segmentation, speech processing, and color classification.

We’re looking for submissions in the following categories:

• Machine Learning on the Edge -- 
  • This refers to systems which connect to the cloud, but are not part of it. Because machine learning involves powerful computers and often powerful GPUs, it is more cost-effective to share resources, and thus cloud computing is a good fit. This is where ML on the edge comes in -- you take a model that may have been trained in the cloud, and then run it on local hardware.
    
• Machine Learning on the Gateway -- 
  • This category refers to running machine learning algorithms on a network gateway device. This has the advantage of a low-latency connection to the cloud, while also being able to perform real-time operations with fairly high throughput. This is a compromise between pure cloud-based machine learning, and pure edge-based machine learning, with some of the benefits and drawbacks of both.
• Artificial Intelligence Blinky -- give Machine Learning a try for the first time by putting it on a simple microprocessor.
• Machine Learning in the Cloud -- 
  • The cloud makes intelligence possible. AWS, Microsoft Azure, and Google Cloud Platform all offer many machine learning options that don't require a deep knowledge of machine learning theory
    

We found a few excellent projects on Hackaday.io 

Using a Raspberry Pi and TensorFlow/Scikit in the cloud, Zach was able to use image analysis to determine which pets should be allowed access to the pet door!

The perfect blackjack robot uses OpenCV and a Raspberry Pi to beat the bank, every time!

An nVidia Jetson gives a batter the edge in knowing if a pitch will be in the strike zone! 

An Arduino 2560 and OpenCV were used to create a prosthetic hand that could intelligently determine the best way to grasp an object.

If you want to study a little bit before diving in, take a look at these courses: 

• MIT Open Courseware
• Introduction to Machine Learning for Coders
• Machine Learning Crash Course

The 4 winners in the 4 categories above will receive $100 Tindie gift certificates (and potentially other prizes!). 

To get started, start a project on hackaday.io and upload your project. 

Contest runs noon January 21 to noon April 7th. All times are in Pacific Standard Time.

How to enter:  Show your project by documenting it as a new project on Hackaday.io. Once you have published your project, look...