Readers are likely familiar with photogrammetry, a method of creating 3D geometry from a series of 2D photos taken of an object or scene. To pull it off you need a lot of pictures, hundreds or even thousands, all taken from slightly different perspectives. Unfortunately the technique suffers where there are significant occlusions caused by overlapping elements, and shiny or reflective surfaces that appear to be different colors in each photo can also cause problems.
But new research from NVIDIA marries photogrammetry with artificial intelligence to create what the developers are calling an Instant Neural Radiance Field (NeRF). Not only does their method require far fewer images, as little as a few dozen according to NVIDIA, but the AI is able to better cope with the pain points of traditional photogrammetry; filling in the gaps of the occluded areas and leveraging reflections to create more realistic 3D scenes that reconstruct how shiny materials looked in their original environment.
If you’ve got a CUDA-compatible NVIDIA graphics card in your machine, you can give the technique a shot right now. The tutorial video after the break will walk you through setup and some of the basics, showing how the 3D reconstruction is progressively refined over just a couple of minutes and then can be explored like a scene in a game engine. The Instant-NeRF tools include camera-path keyframing for exporting animations with higher quality results than the real-time previews. The technique seems better suited for outputting views and animations than models for 3D printing, though both are possible.
At this point, 3D printers are nearly everywhere. Schools, hackerspaces, home workshops, you name it. Most of these machines are of the extruded-filament variety, better known as FDM or Fused Deposition Modelling. Over the last few years, cheap LCD printers have brought resin printing to many shops as well. LCD printers, like their DLP and SLA counterparts, use ultraviolet light to cure liquid resin. These machines are often praised for the super-high detail they can achieve, but are realllly slow. And messy — liquid resin gets everywhere and sticks to everything.
It turns out you can. His extremely well-documented research is shown in the video below, and logs his design process, from initial idea to almost-kinda-working prototype. As you may expect, extruding a high-viscosity liquid at a controlled rate and laser-curing it is not an easy task, but [Jón] made a fantastic attempt. From designing and building his own peristaltic pump, to sending a UV laser through fiber-optic cables, he explored a ton of different approaches to making the printer work. While he may not have been 100% successful, the video is a great reminder that not all projects have to go the way we hope they will.
After you’ve built a snazzy Raspberry Pi-powered retro gaming console, you might be wondering if you could have just a wee bit more power and run some of those other games you might remember, such as Xbox, Wii, or PS3. Perhaps in the future, a later revision of an RPi could handle it but currently, to emulate the 6th/7th generation of consoles, you need something a little beefier. Luckily, [Zac] got his hands on an old gaming laptop and turned it into his own game console.
The first step was to take the laptop apart and discard the parts not needed. [Zac] stripped away the battery, Bluray drive, and spinning hard disk. That left him with a much smaller PCB that could fit into a small case. The power button was integrated into the keyboard but came into the motherboard by the flat cable keyboard connection. So by bridging a few pins, he could power up the laptop. Next, he upgraded the RAM, wifi card, an NVMe drive, and redid all the thermal paste and putty to try and keep things cool while overclocking the GPU.
The case for the machine heavily used his CNC as it was walnut with a mid-section made of plywood. The top has a gorgeous cast acrylic window to see inside. The part the [Zac] was dreading with the fine pitch soldering. Ultimately he got both wires connected with good connections and no bridging. Because it’s just a PC at its heart, almost every game is on the table. Emulation, some more moderate PC games, streaming from his office PC, and cloud gaming services allow him to access most games made. We love the concept and the idea.
Many years of using televisions, monitors, and projectors have conditioned us into treating them as simple peripherals whose cables carry only video. A VGA cable may have an i2c interface for monitor detection, but otherwise it presents little security risk. An HDMI interface on the other hand can carry an increasing number of far more capable ports, meaning that it has made the leap from merely a signal cable to being a connector stuffed with interesting attack vectors for a miscreant. Is it time for an HDMI firewall? [King Kévin] thinks so, because he’s made one.
It’s a surprisingly simple device, because the non-signal capabilities of HDMI rely on a set of conductors which are simply not connected. This of course also disconnects the on-board EEPROM in the device being connected, so there’s an EEPROM on the firewall board to replace it which must be programmed with the information for the device in question.
The premise of HDMI as an attack surface is a valid one, and we’re sure there will be attacks that can be performed on vulnerable displays which could potentially in turn do naughty things to anything which connects to them. The main value for most readers here probably lies though in the introduction it gives to some of what goes into an HDMI interface, and in accessing the i2c interface therein.
Back in 2018, we covered the work being done by [Andrew Sinden] to create a lightgun that could work on modern televisions. The project was looking for funding via Kickstarter, but due at least in part to skepticism about the technology involved, the campaign fell well short of its goal. It seemed, at the time, like the story would end there.
The final version of the hardware ditches the realistic firearm aesthetic inherited from the Wii gun accessory it was designed to fit into, and now features a brightly-colored pistol enclosure that wouldn’t look out of place tethered to a Virtua Cop machine. It’s also gained an optional recoil solenoid for force feedback, though it tacks on another $60 to the already hefty $100 price tag for the base model.
We’re glad to see that [Andrew] recognized the importance of getting Linux support for the software side of things, as it enabled the development of a pre-configured Retropie image for the Raspberry Pi 4. Though you aren’t forced to emulate on the Pi, for those who would like to blast the occasional zombie on their desktop, Windows and x86 Linux are also supported.
Often times, when we cover a project here on Hackaday it’s a one-shot deal: somebody had a particular need or desire, built a gadget to fulfill it, and moved on. There’s nothing wrong with that, but there’s a certain feeling of pride when we see a project from this community develop into something more. While not every hacked together piece of hardware we feature has the potential to be the next Arduboy or Sinden Lightgun, we like to think that we’ve already covered the next big project-turned-product success story and just don’t know it yet.
We are often struck by how often we spend time trying to optimize something when we would be better off just picking a better algorithm. There is the old story about the mathematician Gauss who, when in school, was given busy work to add the integers from 1 to 100. While the other students laboriously added each number, Gauss realized that 100+1 is 101 and 99 + 2 is also 101. Guess what 98 + 3 is? Of course, 101. So you can easily find that there are 50 pairs that add up to 101 and know the answer is 5,050. No matter how fast you can add, you aren’t likely to beat someone who knows that algorithm. So here’s a question: You have a large body of text and you want to search for it. What’s the best way?
We’re big fans of useless machines here at Hackaday, there’s something undeniably entertaining about watching a gadget flail about dramatically without actually making any progress towards a defined goal. But what happens when one of these meme machines ends up working too well? We think that’s just what we might be witnessing here with the Tacobot from [Vije Miller].
On the surface, building an elaborate robotic contraption to (slowly) produce tacos is patently ridiculous. Doubly so when you tack on the need to give it voice commands like it’s some kind of one-dish version of the Star Trek food replicator. The whole thing sounds like the setup for a joke, an assumption that’s only reinforced after watching the dramatized video at the break. But in the end, we still can’t get over how well the thing appears to work.
After [Vije] gives it a list of ingredients to dispense, a robotic arm drops a tortilla on a fantastically articulated rotating platform that can not only spin and move in two dimensions, but can form the soft shell into the appropriate taco configuration. The empty shell is then brought under a rotating dispenser that doles out (or at least attempts to) the requested ingredients such as beef, onions, cheese, and lettuce. With a final flourish, it squirts out a few pumps of the selected sauce, and then presents the completed taco to the user.
The only failing appears to be the machine’s ability to dispense some of the ingredients. The ground beef seems to drop into place without issue, but it visibly struggles with the wetter foodstuffs such as the tomatoes and onions. All we know is that if a robot handed us a taco with that little lettuce on it, we’d have a problem. On the project page [Vije] acknowledges the issue, and says that a redesigned dispenser could help alleviate some of the problem.