Ok Google, Are You a Neural Net?
Neural networks ability to understand the world around us utilizing deep learning has progressed leaps and bounds in the last 5-10 years and what this means for you is, well just about anything you can imagine.
One interesting example of something you can create is Robert Bond’s project where he throws together an autonomous sprinkler system to scare unwanted cats out of his yard!
The story and photographs are quite hilarious and harmless to the neighborhood cats but this project surely gets the job done. We all love cats, just not when they urinate or defecate on our belongings! After his project, Robert compiled and shared all the data on his project with complete instructions which you can find here!
Now you can use this model to build your own sprinkler system encouraging those neighborhood cats to find a new stomping ground. Once they cross into your yard, the sprinklers just turn on and scare them away, simple enough, yet amazingly accurate.
Today, with the necessary tools, you can be a visionary in the realm of deep learning, train your own neural network and show everyone what you’ve got!
Siri, Are You a Neural Net?
Resources You’ll Need:
- Time (We Could All Use a Little More of This)
- Educational Material:
- (Depending on Your Prior Knowledge Going into This Project)
- DIY Deep Learning for Vision: a Hands-On Tutorial with Caffe
- NVIDIA Deep Learning Institute
- Neural Networks for Machine Learning
- Deep Learning for Computer Vision
- Large-Scale Deep Learning for Building Intelligent Computer Systems
- Deep Learning Tutorial by Yann LeCun
- Getting started with Caffe
Lucky for us the internet offers a plethora of educational resources, read on my friend!
So on and so forth, you choose what is right for your desired environment and project
Deep Learning Library/ Software:
- Caffe Deep Learning Framework
- Torch Scientific Computing Framework
- Theano Python Library
- TensorFlow by Google
- Much more!
It’s Time to Hit the Ground Running!
Nvidia offers Jetson Developer Kits in a few different variations including the model TX1 or you can pre-order TX2. These are pretty amazing little joys for the price, great for AI computing, offering guides, instructions, tutorials online and awesome support from the Nvidia developer website. Altogether, you are in good hands.
From there you can find suitable software online, such as the Caffe deep learning vision library. Caffe is a deep learning framework actively in development by the Berkeley Vision and Learning Center in conjunction with community contributors. With this deep learning framework and some Python code, you’ll be up and running in no time!
Here Are a Few Quick Setup Tutorials:
Now once you’ve conceptualized your idea, purchased your gear and downloaded your software it’s time to configure and train your neural network for the task at hand! Deep learning or deep machine learning allows you to train your newly built neural network and you can either start from scratch or you can use transfer learning. Either way, lots and lots of data is not only important, it is necessary. Say you want your machine to classify images of cats and dogs utilizing visual recognition, transfer learning takes work from existing large data sets and cuts time in your process. If you were using Caffe, there is an available repository where researchers and other individuals share their trained models. You can find Model Zoo here.
Also Here Is a Pretty In-Depth Hands-On Introduction:
Remember, what you do with your neural network, what you train it on, is completely up to you. There are plenty of resources available and now you are in control of the product creation and demonstration.