Before 3D printers, there was LEGO. And the little bricks are still useful for putting something together on the quick. Proof is YouTuber [Matthias Wandel]’s awesome bottle cap shooter build that uses rudimentary DIY computer vision to track you and then launch a barrage of plastic pieces at you.
This is an amazing project that has a bit of something for everyone. Lets start with the LEGO. [Matthias Wandel] starts with making a crossbow designed launcher and does an awesome job with showing us how it works in a video. The mechanism is an auto reloading and firing system that can be connected to a stepper motor. Next comes the pan and tilt mechanism which allows the turret to take better aim at moving targets: more LEGO and stepper motors.
The target tracker uses color matching in a program that curiously uses no OpenCV. It compares consecutive frame and then filters out red objects – the largest red dot is it. Since using a fisheye lens on the Raspbery Pi camera adds distortion, [Matthias Wandel] uses a jig made with more Legos to calibrate the image.
The final testing involved having his own child walk around the room being hunted but the autonomous machine. Kids do love toys even if they are trying to shoot bottle caps at them.
Want more Lego inspiration? Check out the Lego Quadcopter Mod and the Lego Tank with the ESP8266.
Continue reading “Wandel Weaponizes Waste With Lego And A Raspberry Pi”
The invention of the blue LED was groundbreaking enough to warrant a Nobel prize. For the last decade, researchers have been trying to take the technology to the next level by controlling the color of emission while the device is in operation. In a new research paper, by the guys over Osaka University, Lehigh University, the University of Amsterdam and West Chester University have presented a GaN LEDs that can be tuned to emit different colors from the same substrate.
GaN or Gallium nitride is a wide band-gap semiconductor that has been employed in the manufacturing of FETs that are known to have higher power density due to its high thermal capacity while increasing efficiency. In the the case of the tunable LED, the key has been the doping with Europium for creating energy bands. When an electron jumps from a higher band to a lower band, it emits energy in the form of light and the wavelength or color depends on the gap of energy jumped as per Plank-Einstein equation.
By controlling the current density and duty cycle, the energy jumps can be controller thereby controlling the color being emitted. This is important since it opens up the possibility of control of LEDs post production. External controllers could be used with the same substrates i.e. same LEDs to make a lamp of different intensity as well as color without needing different doping for R,G and B emissions. The reduction in cost as well as size could be phenomenal and could pave the way for similar semiconductor research.
We have covered the details of the LED in the past along with some fundamentals on the control techniques. We are hoping for some high speed color accurate displays in the future that don’t break the bank on our next gaming build.
Thanks for the tip [Qes]
Adversarial attacks are not something new to the world of Deep Networks used for image recognition. However, as the research with Deep Learning grows, more flaws are uncovered. The team at the University of KU Leuven in Belgium have demonstrated how, by simple using a colored photo held near the torso of a man can render him invisible to image recognition systems based on convolutional neural networks.
Convolutional Neural Networks or CNNs are a class of Deep learning networks that reduces the number of computations to be performed by creating hierarchical patterns from simpler and smaller networks. They are becoming the norm for image recognition applications and are being used in the field. In this new paper, the addition of color patches is seen to confuse the image detector YoLo(v2) by adding noise that disrupts the calculations of the CNN. The patch is not random and can be identified using the process defined in the publication.
This attack can be implemented by printing the disruptive pattern on a t-shirt making them invisible to surveillance system detection. You can read the paper[PDF] that outlines the generation of the adversarial patch. Image recognition camouflage that works on Google’s Inception has been documented in the past and we hope to see more such hacks in the future. Its a new world out there where you hacking is colorful as ever.
Continue reading “The Cloak Of Invisibility Against Image Recognition”
I have a home alarm system that has me wondering if I can make it better with my maker Kung-fu. Recently we had to replace our system, so I took the time to dissect the main controller, the remote sensors, and all the bits that make a home security system work.
To be precise, the subject of today’s interrogation is a Zicom brand Home Alarm that was quite famous a decade ago. It connects to a wired telephone line, takes inputs from motion, door, and gas sensors, and will make quite a racket if the system is tripped (which sometimes happened accidentally). Even though no circuits were harmed in the making of this post, I assure you that there are some interesting things that will raise an eyebrow or two. Lets take a look.
Continue reading “Teardown: The Guts Of A Digital Sentry”
So you just got something like an Arduino or Raspberry Pi kit with a few sensors. Setting up temperature or motion sensors is easy enough. But what are you going to do with all that data? It’s going to need storage, analysis, and summarization before it’s actually useful to anyone. You need a dashboard!
But even before displaying the data, you’re going to need to store it somewhere, and that means a database. You could just send all of your data off into the cloud and hope that the company that provides you the service has a good business model behind it, but frankly the track records of even the companies with the deepest pockets and best intentions don’t look so good. And you won’t learn anything useful by taking the easiest way out anyway.
Instead, let’s take the second-easiest way out. Here’s a short tutorial to get you up and running with a database backend on a Raspberry Pi and a slick dashboard on your laptop or cellphone. We’ll be using scripts and Docker to automate as many things as possible. Even so, along the way you’ll learn a little bit about Python and Docker, but more importantly you’ll have a system of your own for expansion, customization, or simply experimenting with at home. After all, if the “cloud” won’t let you play around with their database, how much fun can it be, really?
Continue reading “Howto: Docker, Databases, And Dashboards To Deal With Your Data”
Good science fiction has sound scientific fact behind it and when Tony Stark first made his debut on the big screen with design tools that worked at the wave of a hand, makers and hackers were not far behind with DIY solutions. Over the years the ideas have become much more polished, as we can see with this Gesture Recognition with PIR sensors project.
The project uses the TPA81 8-pixel thermopile array which detects the change in heat levels from 8 adjacent points. An Arduino reads these temperature points over I2C and then a simple thresholding function is used to detect the movement of the fingers. These movements are then used to do a number of things including turn the volume up or down as shown in the image alongside.
The brilliant part is that the TPA81 8-Pixel sensor has been around for a number of years. It is a bit expensive though it has the ability to detect small thermal variations such as candle flames at up to 2 Meters. More recent parts such as the Panasonic AMG8834 that contain a grid of 8×8 such sensors are much more capable for your hacking/making pleasure, but come with an increased price tag.
This technique is not just limited to gestures, and can be used in Heat-Seeking Robots that can very well be trained to follow the cat around just to annoy it.
AI today is like a super fast kid going through school whose teachers need to be smarter than if not as quick. In an astonishing turn of events, a (satelite)image-to-(map)image conversion algorithm was found hiding a cheat-sheet of sorts while generating maps to appear as it if had ‘learned’ do the opposite effectively[PDF].
The CycleGAN is a network that excels at learning how to map image transformations such as converting any old photo into one that looks like a Van Gogh or Picasso. Another example would be to be able to take the image of a horse and add stripes to make it look like a zebra. The CycleGAN once trained can do the reverse as well, such as an example of taking a map and convert it into a satellite image. There are a number of ways this can be very useful but it was in this task that an experiment at Google went wrong.
A mapping system started to perform too well and it was found that the system was not only able to regenerate images from maps but also add details like exhaust vents and skylights that would be impossible to predict from just a map. Upon inspection, it was found that the algorithm had learned to satisfy its learning parameters by hiding the image data into the generated map. This was invisible to the naked eye since the data was in the form of small color changes that would only be detected by a machine. How cool is that?!
This is similar to something called an ‘Adversarial Attack‘ where tiny amounts of hidden data in an image or other data-set will cause an AI to produce erroneous output. Small numbers of pixels could cause an AI to interpret a Panda as a Gibbon or the ocean as an open highway. Fortunately there are strategies to thwart such attacks but nothing is perfect.
You can do a lot with AI, such as reliably detecting objects on a Raspberry Pi, but with Facial Recognition possibly violating privacy some techniques to fool AI might actually come in handy.