What do you do if you’re an NVIDIA employee and you’re tired of your neighbor’s cats hanging out on your front lawn? Well, if you’re Bob Bond you build a deep learning based controller for your sprinkler system and train it to recognize cats!
Bob’s project uses an IP camera to feed images to a Caffe-based deep learning model pretrained on ImageNet data running on an NVIDIA Jetson TX1 system. This talks to a his sprinkler system to turn on the water when an object identified as a cat or a dog makes its way onto his lawn. Very cool. He’s got a great write-up on his blog and the code is up on github.
Next up is a cool project that shows you how to control a small raspberry pi based robot called a GoPiGo with TensorFlow, to enable simple autonomous driving. There’s a neat video up on YouTube showing the GoPiGo autonomously navigating a simple course and the code that does it is dead simple. In my view, this really shows off the power of deep learning and TensorFlow in particular.
Finally, if you’re looking for more project or research ideas and you’re interested in Natural Language Processing, check out the project reports written by students in Richard Socher’s Deep Learning for NLP course at Stanford, also known as CS224d, which were posted this week. There are some pretty interesting projects including several on sentiment analysis, political bias detection, playlist creation, video annotation, and much more. There appear to be around 100 reports total, so there’s sure to be something for everyone.
