OAK-D Depth Sensing AI Camera Gets Smaller And Lighter

The OAK-D is an open-source, full-color depth sensing camera with embedded AI capabilities, and there is now a crowdfunding campaign for a newer, lighter version called the OAK-D Lite. The new model does everything the previous one could do, combining machine vision with stereo depth sensing and an ability to run highly complex image processing tasks all on-board, freeing the host from any of the overhead involved.

Animated face with small blue dots as 3D feature markers.
An example of real-time feature tracking, now in 3D thanks to integrated depth sensing.

The OAK-D Lite camera is actually several elements together in one package: a full-color 4K camera, two greyscale cameras for stereo depth sensing, and onboard AI machine vision processing with Intel’s Movidius Myriad X processor. Tying it all together is an open-source software platform called DepthAI that wraps the camera’s functions and capabilities together into a unified whole.

The goal is to give embedded systems access to human-like visual perception in real-time, which at its core means detecting things, and identifying where they are in physical space. It does this with a combination of traditional machine vision functions (like edge detection and perspective correction), depth sensing, and the ability to plug in pre-trained convolutional neural network (CNN) models for complex tasks like object classification, pose estimation, or hand tracking in real-time.

So how is it used? Practically speaking, the OAK-D Lite is a USB device intended to be plugged into a host (running any OS), and the team has put a lot of work into making it as easy as possible. With the help of a downloadable application, the hardware can be up and running with examples in about half a minute. Integrating the device into other projects or products can be done in Python with the help of the DepthAI SDK, which provides functionality with minimal coding and configuration (and for more advanced users, there is also a full API for low-level access). Since the vision processing is all done on-board, even a Raspberry Pi Zero can be used effectively as a host.

There’s one more thing that improves the ease-of-use situation, and that’s the fact that support for the OAK-D Lite (as well as the previous OAK-D) has been added to a software suite called the Cortic Edge Platform (CEP). CEP is a block-based visual coding system that runs on a Raspberry Pi, and is aimed at anyone who wants to rapidly prototype with AI tools in a primarily visual interface, providing yet another way to glue a project together.

Earlier this year we saw the OAK-D used in a system to visually identify weeds and estimate biomass in agriculture, and it’s exciting to see a new model being released. If you’re interested, the OAK-D Lite is available at a considerable discount during the Kickstarter campaign.

Intel RealSense D435 Depth Camera

RealSense No Longer Makes Sense For Intel

We love depth-sensing cameras and every neat hack they enabled, but this technological novelty has yet to break through to high volume commercial success. So it was sad but not surprising when CRN reported that Intel has decided to wind down their RealSense product line.

As of this writing, one of the better confirmations for this report can be found on the RealSense SDK GitHub repository README. The good news is that core depth-sensing RealSense products will continue business as usual for the foreseeable future, balanced by the bad news that some interesting offshoots (facial authentication, motion tracking) will be declared “End of Life” immediately and phased out over the next six months.

This information tells us while those living out on the bleeding edge will have to scramble, there is no immediate crisis for everyone else, whether they be researchers, hobbyists, or product planners. But this also means there will be no future RealSense cameras, kicking off many “What’s Next?” discussions in various communities. Like this thread on ROS (Robot Operating System) Discourse.

Three popular alternatives offer distinctly different tradeoffs. The “Been Around The Block” name is Occipital, with their more expensive Structure Pro sensor. The “Old Name, New Face” option is Microsoft Azure Kinect, the latest non-gaming-focused successor to the gaming peripheral that started it all. And let’s not forget OAK-D as the “New Kid On The Block” that started with a crowdfunding campaign and building an user community by doing things like holding contests. Each of these will appeal to a different niche, and we’ll keep our eye open in the future. Let’s see if any of them find the success that eluded the original Kinect, Google’s Tango, and now Intel’s RealSense.

[via Engadget]

Machine-Vision Archer Makes You The Target, If You Dare

We’ll state right up front that it’s a really, really bad idea to let a robotic archer shoot an apple off of your head. You absolutely should not repeat what you’ll see in the video below, and if you do, the results are all on you.

That said, [Kamal Carter]’s build is pretty darn cool. He wisely chose to use just about the weakest bows you can get, the kind with strings that are basically big, floppy elastic bands that shoot arrows with suction-cup tips and are so harmless that they’re intended for children to play with and you just know they’re going to shoot each other the minute you turn your back no matter what you told them. Target acquisition is the job of an Intel RealSense depth camera, which was used to find targets and calculate the distance to them. An aluminum extrusion frame holds the bow and adjusts its elevation, while a long leadscrew and a servo draw and release the string.

With the running gear sorted, [Kamal] turned to high school physics for calculations such as the spring constant of the bow to determine the arrow’s initial velocity, and the ballistics formula to determine the angle needed to hit the target. And hit it he does — mostly. We’re actually surprised how many on-target shots he got. And yes, he did eventually get it to pull a [William Tell] apple trick — although we couldn’t help but notice from his, ahem, hand posture that he wasn’t exactly filled with self-confidence about where the arrow would end up.

[Kamal] says he drew inspiration both from [Mark Rober]’s dart-catching dartboard and [Shane Wighton]’s self-dunking basketball hoop for this build. We’d say his results put in him good standing with the skill-optional sports community.

Continue reading “Machine-Vision Archer Makes You The Target, If You Dare”

New Part Day: Onion Tau LiDAR Camera

The Onion Tau LiDAR Camera is a small, time-of-flight (ToF) based depth-sensing camera that looks and works a little like a USB webcam, but with  a really big difference: frames from the Tau include 160 x 60 “pixels” of depth information as well as greyscale. This data is easily accessed via a Python API, and example scripts make it easy to get up and running quickly. The goal is to be an affordable and easy to use option for projects that could benefit from depth sensing.

When the Tau was announced on Crowd Supply, I immediately placed a pre-order for about $180. Since then, the folks at Onion were kind enough to send me a pre-production unit, and I’ve been playing around with the device to get an idea of how it acts, and to build an idea of what kind of projects it would be a good fit for. Here is what I’ve learned so far.

Continue reading “New Part Day: Onion Tau LiDAR Camera”

OpenCV And Depth Camera Spots Weeds

Using vision technology to identify weeds in agriculture is an area of active development, and a team of researchers recently shared their method of using a combination of machine vision plus depth information to identify and map weeds with the help of OpenCV, the open-source computer vision library. Agriculture is how people get fed, and improving weed management is one of its most important challenges.

Many current efforts at weed detection and classification use fancy (and expensive) multispectral cameras, but PhenoCV-WeedCam relies primarily on an OAK-D stereo depth camera. The system is still being developed, but is somewhat further along than a proof of concept. The portable setups use a Raspberry Pi, stereo camera unit, power banks, an Android tablet for interfacing, and currently require an obedient human to move and point them.

It’s an interesting peek at the kind of hands-on work that goes into data gathering for development. Armed with loads of field data from many different environments, the system can use the data to identify grasses, broad leaf plants, and soil in every image. This alone is useful, but depth information also allows the system to estimate overall plant density as well as try to determine the growth center of any particular plant. Knowing that a weed is present is one thing, but to eliminate it with precision — for example with a laser or mini weed whacker on a robot arm — knowing where the weed is actually growing from is an important detail.

PhenoCV-WeedCam (GitHub repository) is not yet capable of real-time analysis, but the results are promising and that’s the next step. The system currently must be carried by people, but could ultimately be attached to a robotic platform made specifically to traverse fields.

Kinect Gave Us A Preview Of The Future, Though Not The One It Intended

This holiday season, the video game industry hype machine is focused on building excitement for new PlayStation and Xbox consoles. Ten years ago, a similar chorus of hype reached a crescendo with the release of Xbox Kinect, promising to revolutionize how we play. That vision never panned out, but as [Daniel Cooper] of Engadget pointed out in a Kinect retrospective, it premiered consumer technologies that impacted fields far beyond gaming.

Kinect has since withdrawn from the gaming market, because as it turns out gamers are quite content with handheld controllers. This year’s new controllers for a PlayStation or Xbox would be immediately familiar to gamers from ten years ago. Even Nintendo, whose Wii is frequently credited as motivation for Microsoft to develop the Kinect, have arguably taken a step back with Joy-cons of their Switch.

But the Kinect’s success at bringing a depth camera to consumer price levels paved the way to explore many ideas that were previously impossible. The flurry of enthusiastic Kinect hacking proved there is a market for depth camera peripherals, leading to plug-and-play devices like Intel RealSense to make depth-sensing projects easier. The original PrimeSense technology has since been simplified and miniaturized into Face ID unlocking Apple phones. Kinect itself found another job with Microsoft’s HoloLens AR headset. And let’s not forget the upcoming wave of autonomous cars and drones, many of which will see their worlds via depth sensors of some kind. Some might even be equipped with the latest sensor to wear the Kinect name.

Inside the Kinect was also one of the earliest microphone arrays sold to consumers. Enabling the Kinect to figure out which direction a voice is coming from, and isolate it from other noises in the room. Such technology were previously the exclusive domain of expensive corporate conference room speakerphones, but now it forms the core of inexpensive home assistants like an Amazon Echo Dot. Raising the bar so much that hacks needed many more microphones just to stand out.

With the technology available more easily elsewhere, attrition of a discontinued device is reflected in the dwindling number of recent Kinect hacks on these pages. We still see a cool project every now and then, though. As the classic sensor bar itself recedes into history, others will take its place to give us depth sensing and smart audio. But for many of us, Kinect was the ambitious videogame peripheral that gave us our first experience.

Machine Vision Keeps Track Of Grubby Hands

Can you remember everything you’ve touched in a given day? If you’re being honest, the answer is, “Probably not.” We humans are a tactile species, with an outsized proportion of both our motor and sensory nerves sent directly to our hands. We interact with the world through our hands, and unfortunately that may mean inadvertently spreading disease.

[Nick Bild] has a potential solution: a machine-vision system called Deep Clean, which monitors a scene and records anything in it that has been touched. [Nick]’s system uses Jetson Xavier and a stereo camera to detect depth in a scene; he built his camera from a pair of Raspberry Pi cams and a Pi 3B+, but other depth cameras like a Kinect could probably do the job. The idea is to watch the scene for human hands — OpenPose is the tool he chose for that job — and correlate their depth in the scene with the depth of objects. Touch a doorknob or a light switch, and a marker is left on the scene. The idea would be that a cleaning crew would be able to look at the scene to determine which areas need extra attention. We can think of plenty of applications that extend beyond the current crisis, as the ability to map areas that have been touched seems to be generally useful.

[Nick] has been getting some mileage out of that Xavier lately — he’s used it to build an AI umpire and shades that help you find lost stuff. Who knows what else he’ll find to do with them during this time of confinement?

Continue reading “Machine Vision Keeps Track Of Grubby Hands”