Wing Opens The Skies For Drones With UTM

May 20, 2019 by Maya Posch 7 Comments

Yesterday Alphabet (formerly known as Google) announced that their Wing project is launching delivery services per drone in Finland, specifically in a part of Helsinki. This comes more than a month after starting a similar pilot program in North Canberra, Australia. The drone design Wing has opted for consists not of the traditional quadcopter design, but a hybrid plane/helicopter design, with two big propellers for forward motion, along with a dozen small propellers on the top of the dual body design, presumably to give it maximum range while still allowing the craft to hover.

With a weight of 5 kg and a wingspan of about a meter, Wing’s drones are capable of lifting and carrying a payload of about 1.5 kg. This puts it into a category of drones far beyond of what hobbyists tend to fly on a regular basis, and worse, it involves Beyond Visual Line Of Sight (BVLOS for short) flying, which is frowned upon by the FAA and similar regulatory bodies. What Google/Alphabet figures that can enable them to make this kind of service a commercial reality is called Unmanned aircraft system Traffic Management (UTM).

UTM is essentially complementary to the existing air traffic control systems, allowing drones to integrate into these flows of manned airplanes without endangering either. Over the past years, it’s been part of NASA’s duty to develop the systems and infrastructure that would be required to make UTM a reality. Working together with the FAA and companies such as Amazon and Alphabet, the hope is that before long it’ll be as normal to send a drone into the skies for deliveries and more as it is today to have passenger and cargo planes with human pilots take to the skies.

Google Launches AI Platform That Looks Remarkably Like A Raspberry Pi

March 5, 2019 by Brian Benchoff 80 Comments

Google has promised us new hardware products for machine learning at the edge, and now it’s finally out. The thing you’re going to take away from this is that Google built a Raspberry Pi with machine learning. This is Google’s Coral, with an Edge TPU platform, a custom-made ASIC that is designed to run machine learning algorithms ‘at the edge’. Here is the link to the board that looks like a Raspberry Pi.

This new hardware was launched ahead of the TensorFlow Dev Summit, revolving around machine learning and ‘AI’ in embedded applications, specifically power- and computationally-limited environments. This is ‘the edge’ in marketing speak, and already we’ve seen a few products designed from the ground up to run ML algorithms and inference in embedded applications. There are RISC-V microcontrollers with machine learning accelerators available now, and Nvidia has been working on this for years. Now Google is throwing their hat into the ring with a custom-designed ASIC that accelerates TensorFlow. It just so happens that the board looks like a Raspberry Pi.

Continue reading “Google Launches AI Platform That Looks Remarkably Like A Raspberry Pi” →

Hack Your Gmail: A Quick Start For Google App Scripting

January 17, 2019 by Al Williams 18 Comments

For many people, Gmail is synonymous with e-mail. Some people like having cloud access to everything and some people hate having any personal data in the cloud. However you feel about it, one thing that was nice about having desktop software is that you could hack it relatively easily. If you didn’t like how your desktop mail client worked, you had a lot of options: use a different program, write your own, hack the executable of your current program, or in the case of open source just fork it and make any changes you are smart enough to make.

Google provides a lot of features with all of its products, but however you slice it, all the code runs on their servers out of your reach. Sort of. If you know JavaScript, you can use Google Apps Script to add features to many Google products including Gmail. If you’ve used Office scripting, the idea is the same, although obviously the implementation is very different.

With scripting you can make sophisticated filters that would be very hard to do otherwise. For example, monitor for suspicious messages like those with more than 4 attachments, or that appear to come from a contact between the hours of 2AM and 5AM.

For our example today, I’m going to show you something that is easy but also highly useful.

Continue reading “Hack Your Gmail: A Quick Start For Google App Scripting” →

Improving Depth Of Field With Only 5 Phones

December 11, 2018 by Lewin Day 46 Comments

The hottest new trend in photography is manipulating Depth of Field, or DOF. It’s how you get those wonderful portraits with the subject in focus and the background ever so artfully blurred out. In years past, it was achieved with intelligent use of lenses and settings on an SLR film camera, but now, it’s all in the software.

The franken-camera rig, consisting of five Pixel 3 smartphones. The cameras are synchronised over WiFi.

For the Pixel 2 smartphone, Google had used some tricky phase-detection autofocus (PDAF) tricks to compute depth data in images, and used this to decide which parts of images to blur. Distant areas would be blurred more, while the subject in the foreground would be left sharp.

This was good, but for the Pixel 3, further development was in order. A 3D-printed phone case was developed to hold five phones in one giant brick. The idea was to take five photos of the same scene at the same time, from slightly different perspectives. This was then used to generate depth data which was fed into a neural network. This neural network was trained on how the individual photos relate to the real-world depth of the scene.

With a trained neural network, this could then be used to generate more realistic depth data from photos taken with a single camera. Now, machine learning is being used to help your phone decide which parts of an image to blur to make your beautiful subjects pop out from the background.

Comparison images show significant improvement of the “learned” depth data versus just the stereo-PDAF generated depth data. It’s yet another shot fired in the smartphone camera arms race, that shows no signs of abating. We just wonder when the Geiger counter mods are going to ship from factory.

[via AndroidPolice]

Piano Genie Trained A Neural Net To Play 88-Key Piano With 8 Arcade Buttons

October 27, 2018 by Steven Dufresne 5 Comments

Want to sound great on a Piano using only your coding skills? Enter Piano Genie, the result of a research project from Google AI and DeepMind. You press any of eight buttons while a neural network makes sure the piano plays something cool — compensating in real time for what’s already been played.

Almost anyone new to playing music who sits down at a piano will produce a sound similar to that of a cat chasing a mouse through a tangle of kitchen pots. Who can blame them, given the sea of 88 inexplicable keys sitting before them? But they’ll quickly realize that playing keys in succession in one direction will produce sounds with consistently increasing or decreasing pitch. They’ll also learn that pressing keys for different lengths of times can improve the melody. But there’s still 88 of them and plenty more to learn, such as which keys will sound harmonious when played together.

With Pinao Genie, gone are the daunting 88 keys, replaced with a 3D-printed box of eight arcade-style buttons which they made by following this Adafruit tutorial. A neural network maps those eight buttons to something meaningful on the 88-key piano keyboard. Being a neural network, the mapping isn’t a fixed one-to-one or even one-to-many. Instead, it’s trained to play something which should sound good taking into account what was play previously and won`t necessarily be the same each time.

To train it they use data from the approximately 1400 performances of the International Piano e-Competition. The result can be quite good as you can see and hear in the video below. The buttons feed into a computer but the computer plays the result on an actual piano.

For training, the neural network really consists of two networks. One is an encoder, in this case a recurrent neural network (RNN) which takes piano sequences and learns to output a vector. In the diagram, the vector is in the middle and has one element for each of the eight buttons. The second network is the decoder, also an RNN. It’s trained to turn that eight-element vector back into the same music which was fed into the encoder.

Once trained, only the decoder is used. The eight-button keyboard feeds into the vector, and the decoder outputs suitable notes. The fact that they’re RNNs means that rather than learning a fixed one-to-many mapping, the network takes into account what was previously played in order to come up with something which hopefully sounds pleasing. To give the user a little more creative control, they also trained it to realize when the user is playing a rising or falling melody and to output the same. See their paper for how the turned polyphonic sound into monophonic and back again.

If you prefer a different style of music you can train it on a MIDI collection of your own choosing using their open-sourced model. Or you can try it out as is right now through their web interface. I’ll admit, I started out just banging on it, producing the same noise I would get if I just hammered away randomly on a piano. Then I switched to thinking of making melodies and the result started sounding better. So some music background and practice still helps. For the video below, the researcher admits to having already played for a few hours.

This isn’t the first project we’ve covered by these Google researchers. Another was this music synthesizer again using neural networks but this time with a Raspberry Pi. And if our discussion of recurrent neural networks went a bit over your head, check out our overview of neural networks.

Continue reading “Piano Genie Trained A Neural Net To Play 88-Key Piano With 8 Arcade Buttons” →

Google Discovers Google+ Servers Are Still Running

October 9, 2018 by Tom Nardi 70 Comments

Google is pulling the plug on their social network, Google+. Users still have the better part of a year to say their goodbyes, but if the fledgling social network was a ghost town before, news of its imminent shutdown isn’t likely to liven the place up. A quick check of the site as of this writing reveals many users are already posting their farewell messages, and while there’s some rallying behind petitions to keep the lights on, the majority realize that once Google has fallen out of love with a project there’s little chance of a reprieve.

To say that this is a surprise would be disingenuous. We’d wager a lot of you already thought it was gone, honestly. It’s no secret that Google’s attempt at a “Facebook Killer” was anything but, and while there was a group of dedicated users to be sure, it never attained anywhere near the success of its competition.

According to a blog post from Google, the network’s anemic user base isn’t the only reason they’ve decided to wind down the service. A previously undisclosed security vulnerability also hastened its demise, a revelation which will particularly sting those who joined for the privacy-first design Google touted. While this fairly transparent postmortem allows us to answer what ended Google’s grand experiment in social networking, there’s still one questions left unanswered. Where are the soon to be orphaned Google+ users supposed to go?

Continue reading “Google Discovers Google+ Servers Are Still Running” →