Lorem ipsum dolor sit amet, consectetur adicing elit ut ullamcorper. leo, eget euismod orci. Cum sociis natoque penati bus et magnis dis.Proin gravida nibh vel velit auctor aliquet. Leo, eget euismod orci. Cum sociis natoque penati bus et magnis dis.Proin gravida nibh vel velit auctor aliquet.

  /  Project   /  Blog: Heartbeat Newsletter Vol. 47

Blog: Heartbeat Newsletter Vol. 47

AI-powered 3D home tours, Edge TPU vs Jetson Nano, an OpenCV-Python cheat sheet, and more

We love putting this newsletter together each week, and we want you to love it, too. If you can, take a few minutes to complete this survey to let us know how we’re doing and what you’d like to see moving forward. Thank you for your time, and happy reading!

~ The Heartbeat Editors


Zillow launches AI-powered 3D home tours on mobile app

3D home tours used to be reserved for high-end real estate, but with the release of their iOS app 3D Home, Zillow is offering 360-degree photos and tours for a much larger range of properties across the U.S. and Canada. In the app, on-screen icons instruct users where and when to snap photos. Once the entire house has been captured, computer vision models adjust exposure levels, select thumbnail images for each room, and stitch together 3D walkthroughs. [Read More]

Amazon Alexa: Using wake word acoustics to filter out background speech improves speech recognition by 15%

One of the most complicated problems in improving voice assistant performance is teaching them to ignore speech not intended for them. The Amazon Alexa team is introducing a new technique to do just this. Essentially, this new technique uses an acoustic snapshot of the wake word and compares subsequent speech to it. When subsequent acoustics match those of the wake word, that speech is recognized as being intended for Alexa — and other speech is treated as background noise. A good technical look at how this is possible. [Read More]

Uber’s autonomous vehicle unit collects $1 billion investment

Japanese conglomerate SoftBank’s Vision Fund, Toyota, and component supplier Denso have collectively invested 10 figures in Uber Advanced Technologies Group, Uber’s self-deriving car wing. This investment comes ahead of Uber’s IPO, which is expected raise around $10 billion, putting the company’s valuation at $100 billion. The investment certainly puts Uber at the forefront of the autonomous vehicle industry, and it will be interesting to see how or if this changes their roadmap (pun intended). [Read More]

Improving security as AI moves to smartphones

Using quantization to compress machine learning models for mobile has gained significant momentum, as major platforms (Core ML 2, TensorFlow, etc.) have baked quantization into their workflows, and models that were once too heavy for edge deployment can now power AI mobile experiences. But in a new study, researchers from MIT and IBM show that compressed models can be quite vulnerable to adversarial attacks. The researchers do offer a fix, however: add a mathematical constraint during the quantization process to reduce the chance that a model will fall prey to a slightly modified image and misclassify what it sees. [Read More]


sobri909 / LocoKit

A Machine Learning based location recording and activity detection framework for iOS. [Explore]

tensorflow / magenta-js

Magenta.js: Music and Art Generation with Machine Intelligence in the Browser. [Explore]

doc-ai / tensorio-ios

Objective-C and Swift wrappers for TensorFlow and TensorFlow Lite. [Explore]


Increasing the Accuracy of the Machine Learning Model in the CarLens Mobile App

Part two of NetGuru’s deep dive into the creation of CarLens, a mobile app that allows users to identify cars on the street and view information about them through an AR layer. [Read More]

Google Coral Edge TPU vs NVIDIA Jetson Nano

A performance comparison of these two new edge devices. Includes performance visualizations and a great analysis of each piece of hardware. [Read More]

[Video] 70fps real-time object detection with Google’s Coral Dev Board with Edge TPU

Here’s a slightly different look at the Edge TPU’s performance. Included is a stream of real-time object detection and a link to other demos. [Read More]

AutoML: The Next Wave of Machine Learning

This post defines AutoML, explores its potential applications, and provides an overview of 7 useful tools for automating machine learning pipelines. [Read More]

OpenCV-Python Cheat Sheet: From Importing Images to Face Detection

Cropping, resizing, rotating, thresholding, blurring, drawing & writing on an image, face detection, & contouring to detect objects. All explained. [Read More]

Source: Artificial Intelligence on Medium

(Visited 3 times, 1 visits today)
Post a Comment