There was a time when making a machine to identify objects in a camera was difficult, even without trying to do it in real time. But now, you can do it with a Jetson Nano board for under $60. How well does it work? Watch [Murtaza’s] video below and see what you think.
The first few minutes of the video piqued our interest, and good thing, too, because the 50 lines of code get a 50-plus minute video! It is worth watching, though, because there’s a lot of good information about how to apply this technique in your own projects.
If this sounds familiar, it’s because he tried the same trick with the RTX 2070 back in January but couldn’t get it working. When he first published the video showing the process of desoldering the 3070’s eight Hynix 1 GB memory chips and replacing them with eight Samsung 2 GB chips he hit the same wall — the card would boot and detect the increased RAM, but was unstable and would eventually crash. Helpful hints from his viewers led him to use an EVGA configuration GUI to lock the operating frequency which fixed the problem. Further troubleshooting (YouTube comment in Russian and machine translation of it) showed that the “max performance mode” setting in the Nvidia tool is also a solution to stabilize performance.
The new memory chips don’t self-report their specs to the configuration tool. Instead, a set of three resistors are used to electronically identify which hardware is present. The problem was that [VIK-on] had no idea which resistors and what the different configurations accomplished. It sounds like you can just start changing zero Ohm resistors around to see the effect in the GUI, as they configure both the brand of memory and the size available. The fact that this board is not currently sold with a 16 GB option, yet the configuration tool has settings for it when the resistors are correctly configured is kismet.
So did it make a huge difference? That’s difficult to say. He’s running some benchmarks in the video, both Unigine 2 SuperPosition and 3DMark Time Spy results are shown. However, we didn’t see any tests run prior to the chip swap. This would have been the key to characterizing the true impact of the hack. That said, reworking these with a handheld hot air station, and working your way through the resistor configuration is darn impressive no matter what the performance bump ends up being.
We all seem to intuitively know that a lot of what we do online is not great for our mental health. Hang out on enough social media platforms and you can practically feel the changes your mind inflicts on your body as a result of what you see — the racing heart, the tight facial expression, the clenched fists raised in seething rage. Not on Hackaday, of course — nothing but sweetness and light here.
That’s all highly subjective, of course. If you’d like to quantify your online misery more objectively, take a look at the aptly named BrowZen, a machine learning application by [Nick Bild]. Built around an NVIDIA Jetson Xavier NX and a web camera, BrowZen captures images of the user’s face periodically. The expression on the user’s face is classified using a facial recognition model that has been trained to recognize facial postures related to emotions like anger, surprise, fear, and happiness. The app captures your mood and which website you’re currently looking at and stores the results in a database. Handy charts let you know which sites are best for your state of mind; it’s not much of a surprise that Twitter induces rage while Hackaday pushes [Nick]’s happiness button. See? Sweetness and light.
Most of us make do with the VRAM that came with our graphics cards. We can just wait until the next one comes out and get a little more memory. After all, it’d be madness to try and delicately solder on new components of something so timing-sensitive as RAM chips, right?
[VIK-on] took it upon himself to do just that. The inspiration came when a leaked diagram suggested that the RTX 2000 line could support 16 GB of RAM by using 2GB chips. NVIDIA never did release a 16GB version of the 2070, so this card is truly one of a kind. After some careful scouring of the internet, the GDDR6 chips were procured and carefully soldered on with a hot air gun. A few resistors had to be moved to accommodate the new RAM chips. During power-on, [VIK-on] saw all 16 GB enumerate and was able to run some stress tests. Unfortunately, the card wasn’t stable and started having black screen issues and wonky clocks. Whether it was a bad solder joint or firmware issues, it’s hard to say but he is pretty convinced it is a BIOS error. Switching the resistors back to the 8GB configuration yielded a stable system.
With the Jetson Nano, NVIDIA has done a fantastic job of bringing GPU-accelerated machine learning to the masses. For less than the cost of a used graphics card, you get a turn-key Linux computer that’s ready and able to handle whatever AI code you throw at it. But if you’re trying to set up a lab for 30 students, the cost of even relatively affordable development boards can really add up.
Spoiler: These things don’t exist.
Which is why [Tea Vui Huang] has developed jetson-emulator. This Python library provides a work-alike environment to NVIDIA’s own “Hello AI World” tutorials designed for the Jetson family of devices, with one big difference: you don’t need the actual hardware. In fact, it doesn’t matter what kind of computer you’ve got; with this library, anything that can run Python 3.7.9 or better can take you through NVIDIA’s getting started tutorial.
So what’s the trick? Well, if you haven’t guessed already, it’s all fake. Obviously it can’t actually run GPU-accelerated code without a GPU, so the library [Tea] has developed simply pretends. It provides virtual images and even “live” camera feeds to which randomly generated objects have been assigned.
The original NVIDIA functions have been rewritten to work with these feeds, so when you call something like net.Classify(img) against one of them you’ll get a report of what faux objects were detected. The output will look just like it would if you were running on a real Jetson, down to providing fictitious dimensions and positions for the bounding boxes.
If you’re a hacker looking to dive into machine learning and computer vision, you’d be better off getting a $59 Jetson Nano and a webcam. But if you’re putting together a workshop that shows a dozen people the basics of NVIDIA’s AI workflow, jetson-emulator will allow everyone in attendance to run code and get results back regardless of what they’ve got under the hood.
Over the past decade, we’ve seen great strides made in the area of AI and neural networks. When trained appropriately, they can be coaxed into generating impressive output, whether it be in text, images, or simply in classifying objects. There’s also much fun to be had in pushing them outside their prescribed operating region, as [Jon Warlick] attempted recently.
[Jon]’s work began using NVIDIA’s GauGAN tool. It’s capable of generating pseudo-photorealistic images of landscapes from segmentation maps, where different colors of a 2D image represent things such as trees, dirt, or mountains, or water. After spending much time toying with the software, [Jon] decided to see if it could be pressed into service to generate video instead.
The GauGAN tool is only capable of taking in a single segmentation map, and outputting a single image, so [Jon] had to get creative. Experiments were undertaken wherein a video was generated and exported as individual frames, with these frames fed to GauGAN as individual segmentation maps. The output frames from GauGAN were then reassembled into a video again.
The results are somewhat psychedelic, as one would expect. GauGAN’s single image workflow means there is only coincidental relevance between consecutive frames, creating a wild, shifting visage. While it’s not a technique we expect to see used for serious purposes anytime soon, it’s a great experiment at seeing how far the technology can be pushed. It’s not the first time we’ve seen such technology used to create full motion video, either. Video after the break.
NVIDIA kicked off their line of GPU-accelerated single board computers back in 2014 with the Jetson TK1, a $200 USD development system for those looking to get involved with the burgeoning world of so-called “edge computing”. It was designed to put high performance computing in a small and energy efficient enough package that it could be integrated directly into products, rather than connecting to a data center half-way across the world.
The TK1 was an impressive piece of hardware, but not something the hacker and maker community was necessarily interested in. For one thing, it was fairly expensive. But perhaps more importantly, it was clearly geared more towards industry types than consumers. We did see the occasional project using the TK1 and the subsequent TX1 and TX2 boards, but they were few and far between.
Then came the Jetson Nano. Its 128 core Maxwell CPU still packed plenty of power and was fully compatible with NVIDIA’s CUDA architecture, but its smaller size and $99 price tag made it far more attractive for hobbyists. According to the company’s own figures, the number of active Jetson developers has more than tripled since the Nano’s introduction in March of 2019. With the platform accessible to a larger and more diverse group of users, new and innovative applications for machine learning started pouring in.
Cutting the price of the entry level Jetson hardware in half was clearly a step in the right direction, but NVIDIA wanted to bring even more developers into the fray. So why not see if lightning can strike twice? Today they’ve officially announced that the new Jetson Nano 2GB will go on sale later this month for just $59. Let’s take a close look at this new iteration of the Nano to see what’s changed (and what hasn’t) from last year’s model.
