Google’s Inception Sees This Turtle As A Gun; Image Recognition Camouflage

November 3, 2017 by 52 Comments

The good people at MIT’s Computer Science and Artificial Intelligence Laboratory [CSAIL] have found a way of tricking Google’s InceptionV3 image classifier into seeing a rifle where there actually is a turtle. This is achieved by presenting the classifier with what is called ‘adversary examples’.

Adversary examples are a proven concept for 2D stills. In 2014 [Goodfellow], [Shlens] and [Szegedy] added imperceptible noise to the image of a panda that from then on was classified as gibbon. This method relies on the image being undisturbed and can be overcome by zooming, blurring or rotating the image.

The applicability for real world shenanigans has been seriously limited but this changes everything. This weaponized turtle is a color 3D print that is reliably misclassified by the algorithm from any point of view. To achieve this, some knowledge about the classifier is required to generate misleading input. The image transformations, such as rotation, scaling and skewing but also color corrections and even print errors are added to the input and the result is then optimized to reliably mislead the algorithm. The whole process is documented in [CSAIL]’s paper on the method.

What this amounts to is camouflage from machine vision. Assuming that the method also works the other way around, the possibility of disguising guns (or anything else) as turtles has serious implications for automated security systems.

As this turtle targets the Inception algorithm, it should be able to fool the DIY image recognition talkbox that Hackaday’s own [Steven Dufresne] built.

Thanks to [Adam] for the tip.

Yellowing: The Plastic Equivalent Of A Sunburn

November 3, 2017 by 62 Comments

Your fancy white electronic brick of consumer electronics started off white, but after some time it yellowed and became brittle. This shouldn’t have happened; plastic is supposed to last forever. It turns out that plastic enclosures are vulnerable to the same things as skin, and the effects are similar. When they are stared at by the sun, the damage is done even though it might not be visible to you for quite some time.

Continue reading “Yellowing: The Plastic Equivalent Of A Sunburn” →

ESP8266 As A Tape Drive

November 3, 2017 by 25 Comments

1976 was the year the Apple I was released, one of several computers based on the MOS 6502 chip. MOS itself released the KIM-1 (Keyboard Input Monitor) initially to demonstrate the power of the chip. The single board computer had two connectors on it, one of which could be used for a tape recorder for long-term storage. When [Willem Aandewiel] went to the Apple Museum Nederland in 2016, he saw one and felt nostalgic for his youth. He was able to get a replica, the microKIM, and build it but he wanted to use new technology to interface with this old technology, so he decided to use an ESP8266 as a solid state tape recorder.

One of the reasons the KIM-1 was so popular when it was released was that there was lots of documentation available. [Willem] used this documentation to figure out how the KIM-1 saves data to the recording device. An ATTiny85 is used to decode the pulse stream that the KIM-1 sends when saving because the timing was too tight to both “listen” and decode the bits as well as convert and store them. For loading programs, the data can be sent digitally as 1’s and 0’s to the KIM-1. This means that the ATTiny is only used for decoding and doesn’t have to re-encode the data.  Because of this, saving is slow, but loading is very quick.

To complete the project, [Willem] added four buttons, one each for rewind, record, play and fast-forward, and a screen so you can see which program is currently selected and can go from one program to another. As a nice throwback touch, record and play have to be pressed at the same time when saving. For more 6502 projects, check out this 6502 based DIY computer, or this 6502 built from discrete parts.

Continue reading “ESP8266 As A Tape Drive” →

Meccano Piston Pump Made With A Syringe

November 3, 2017 by 8 Comments

[Mohamed Sami] built a syringe pump out of Meccano building set parts. It consists of a simple framework with a DC motor mounted on it that actuates the syringe when powered. A check valve harvested from an ordinary household spray bottle keeps the syringe from sucking back liquid that it has just pumped out, so it can keep pumping forever. A lead-acid battery powers the whole thing.

Syringe pumps are typically used to deliver precisely measured quantities of substances. Right now [Mohamed]’s rig is just an uncontrolled pump, but he hopes to get a better understanding of and control over how much liquid gets pumped. Adding an encoder to the DC motor would be a start, was his thought — or even better would be a stepper.

You’d be surprised how many syringe pump projects we publish. Not just another syringe pump, but simple hydraulic projects and even using the syringe barrel as a logic probe’s enclosure.

Building A Monster Floodlight Out Of Scrap

November 2, 2017 by 29 Comments

When the apocalypse comes, we want [Justin] on our team! He made a hefty 400 W work light out of four 100 W LEDs mounted to a giant, aluminum slab-like heat sink he had lying around. He manufactured a diffuser for the LEDs by cutting down what appeared to be a old broken fluorescent light fixture’s cover, with side plates bolted into place for good measure. [Justin] does a lot of metalwork in his projects, and you can see it the precision with which he bolts the various parts together into a rather slick assembly.

The LEDs run off 110 V, and [Justin] soldered one of those white iPhone USB chargers in to power four small fans that are mounted on the heat sink fins backing up the LEDs. Then he mounted a ball joint onto the underside so the thing could be pointed wherever, with the other end of the joint attached to what might be the tripod from a standing work light.

Now all he needs is a control system, like this arcade button workshop light or this fully controllable workshop lighting rig.

Continue reading “Building A Monster Floodlight Out Of Scrap” →

Creating Modular Storage Out Of Used Filament Spools

November 2, 2017 by 20 Comments

[Alec Richter] had a good idea on how he could convert the leftover filament spindles from his 3D printer into multi-compartment storage. An empty spindle is fitted with several trays that rotate out from the circle for easy access. With multiple spools rotating on a central axle, you can really see how a bunch of parts could be organized in a column, though not being able to see through the sides probably limits its use somewhat — most of the modular component storage we’ve seen has clear trays.

He has designed drawer bases with removable compartment trays, along with alignment jigs to help you get the drawer installed perfectly the first time. You can download the designs (14 files!) but you need to sign up for an account first. Also, [Alex]s designs fit very specific spindles so be sure of your measurements, etc.

Hackaday is awash in posts about modular storage, like this computer tower turned storage shelf and this technique for using peanut butter jars for storage.

[mucho apreciado for the tip, George!]

Why Wait? Just Plate Your Own PCB Vias

November 2, 2017 by 34 Comments

[Jan Mrázek] is a pro when it comes to rolling his own PCBs. He can crank out a 6/6 mil double-sided PCB in 45 minutes flat. As a challenge to his prowess, he decided to experiment with plating through-hole PCBs at home, because sometimes you just can’t wait for China to deliver the goods.

The key here is to make a non-conductive surface—the walls of holes drilled in a sheet of copper clad–conductive. While there are some established ways of doing this at home, the chemicals are difficult to source. When his local supplier started stocking colloidal graphite paint, which is used to prevent ESD and fix non-working remote control buttons, he decided to try it.

[Jan] drilled up a board with holes ranging from 0.1mm up to 8mm, polished it, and gave it an acetone bath. He sprayed each side with graphite and cured it at 100 °C for 20 minutes. At this point, wall hole resistance measured 21 Ω. [Jan] wet-sanded away the graphite and set up an electroplating bath. Right away, he could see a layer of copper forming on the holes. After 90 minutes, he polished the board again and separated the vias to prepare for the real test: solder. This time, every hole except the smallest size reported a resistance of 0.1 Ω. But they all sucked solder through the vias, making this experiment a success.

[Jan] concluded that this is a simple and effective process, but is rarely worth the effort. We wonder how the simplicity of this method compares to drilling wells instead of holes, filling them with conductive ink, and then drilling the rest of the via.

Via [Dangerous Prototypes]