We’ve all come to terms with a neural network doing jobs such as handwriting recognition. The basics have been in place for years and the recent increase in computing power and parallel processing has made it a very practical technology. However, at the core level it is still a digital computer moving bits around just like any other program. That isn’t the case with a new neural network fielded by researchers from the University of Wisconsin, MIT, and Columbia. This panel of special glass requires no electrical power, and is able to recognize gray-scale handwritten numbers.
As much as we’d like to have the right tools for the right job all of the time, sometimes our parts drawers have other things in mind. After all, what’s better than buying a new tool than building one yourself from things you had lying around? That’s at least what [Saulius] must have been thinking when he needed a thermometer with a digital output, but only had a dumb, but feature-rich, thermometer on hand.
Luckily, [Saulius] had a webcam lying around as well as an old thermometer, and since the thermometer had a LCD display it was relatively straightforward to get the camera to recognize the digits in the thermometer’s display. This isn’t any old thermometer, either. It’s a four-channel thermometer with good resolution and a number of other useful features (with an obvious lack of communications abilities), so it’s not something that he could just overlook.
Once the camera was mounted to an arm and pointed at the thermometer’s screen, an algorithm running on a computer detects polygons and reports its information into a CSV file. This process is made simpler by the fact that LCD screens like this are very predictable. From there, the data is imported into LibreOffice and various charts and graphs can be made.
Although perhaps not the most elegant of hacks, sometimes you have to work with the supplies that are on hand at the time. Sometimes the tools you need are too expensive, politically dangerous, or too impractical to obtain. To that end [Saulius]’s hack is a great example of what hacks are possible with the right mindset.
We are all (hopefully) aware that we can be watched while we’re online. Our clicks are all trackable to some extent, whether it’s our country’s government or an advertiser. What isn’t as obvious, though, is that it’s just as easy to track our movements in real life. [Saulius] was able to prove this concept by using optical character recognition to track the license plate numbers of passing cars half a kilometer away.
To achieve such long distances (and still have clear and reliable data to work with) [Saulius] paired a 70-300 mm telephoto lens with a compact USB camera. All of the gear was set up on an overpass and the camera was aimed at cars coming around a corner of a highway. As soon as the cars enter the frame, the USB camera feeds the information to a laptop running openALPR which is able to process and record license plate data.
The build is pretty impressive, but [Saulius] notes that it isn’t the ideal setup for processing a large amount of information at once because of the demands made on the laptop. With this equipment, monitoring a parking lot would be a more feasible situation. Still, with even this level of capability available to anyone with the cash, imagine what someone could do with the resources of a national government. They might even have long distance laser night vision!