Computer vision promises to revolutionize everything from smartphone security to document scanning, but the concepts aren't the easiest to wrap your head around. To help budding data scientists get experimenting, Google, one of the field's foremost authorities, is releasing a machine learning development kit for computer vision. Google says it's the first device to bring offline, on-device neural network acceleration to consumers.
Google's AIY Vision Kit, which will be available for purchase later this month, comes with most (but not all) of what you need to get programming. It includes a circuit board, a computer vision kit software, a speaker, RGB lights, and a cardboard box to house all of the hardware. Unfortunately, you'll have to buy the camera — specifically the Raspberry Pi Camera, which is designed to play nicely with the Raspberry Pi's hardware — separately.
The kit uses a VisionBonnet board for Raspberry Pi, a low-power visual processing unit powered by Google's TensorFlow AI platform. And it ships with three pre-trained computer vision models: MobileNets, which recognizes up to a thousand common objects; a facial recognition and expression detection algorithm that distinguishes between people; and a pet-detecting model that picks out humans, cats, and dogs in the camera's field of view.
Optionally, you can deploy your own TensorFlow models and retrain them, or use the AIY's Python application programming interface (API) to customize parameters. If you're in need of ideas, Google's supplied a few suggestions:
- Identify all kinds of plant and animal species
- See when your dog is at the back door
- See when your car left the driveway
- See that your guests are delighted by your holiday decorations
- See when your little brother comes into your room
The kit is a build-it-yourself affair, and it'll cost $45 when it ships on December 31.
It's the second AIY kit to come out of Google's machine learning labs, the first being the search giant's AIY Voice Kit. It, much like the AIY Vision Kit, let hobbyists play with machine learning using a Raspberry Pi and a few extra components.
Looking to get started? Check out the Google Blog post down below.
Source: Google Blog
from xda-developers http://ift.tt/2AxSwPe
via IFTTT
Aucun commentaire:
Enregistrer un commentaire