Explore Neural Radiance Fields In Real-time, Even On A Phone

December 20, 2023 by Donald Papp 8 Comments

Neural Radiance Fields (NeRF) is a method of reconstructing complex 3D scenes from sparse 2D inputs, and the field has been growing by leaps and bounds. Viewing a reconstructed scene is still nontrivial, but there’s a new innovation on the block: SMERF is a browser-based method of enabling full 3D navigation of even large scenes, efficient enough to render in real time on phones and laptops.

Don’t miss the gallery of demos which will run on anything from powerful desktops to smartphones. Notable is the distinct lack of blurry, cloudy, or distorted areas which tend to appear in under-observed areas of a NeRF scene (such as indoor corners and ceilings). The technical paper explains SMERF’s approach in more detail.

NeRFs as a concept first hit the scene in 2020 and the rate of advancement has been simply astounding, especially compared to demos from just last year. Watch the short video summarizing SMERF below, and marvel at how it compares to other methods, some of which are themselves only months old.

Continue reading “Explore Neural Radiance Fields In Real-time, Even On A Phone” →

Terminal-Based Image Viewer, and Multi-OS Binary, and Under 100kb

December 11, 2023 by Donald Papp 7 Comments

[Justine Tunney]’s printimage.com is a program capable of splatting full-color images to text mode terminal sessions, but that’s not even its neatest trick. It’s also a small binary executable capable of running on six different operating systems: Linux, Windows, MacOS, FreeBSD, OpenBSD, and NetBSD. All without having to be installed or otherwise compiled first. On top of it all, it’s less than 100 kb.

How is this possible? It’s thanks to [Justine]’s αcτµαlly pδrταblε εxεcµταblε format, implemented by a project called Cosmopolitan which aims to turn C into a build-once-run-anywhere language. The printimage.com source code is included within the Cosmopolitan project.

If the name sounds a bit familiar, it’s probably because the Cosmopolitan project is a key piece of a tool we recently covered: llamafile, which allows people to package up an LLM (large language model) as a single-file, multi-OS executable.

As printimage.com shows, terminal windows are capable of more than just text. Still, plain ASCII has its appeal. Check out the ASCII art STL file viewer which might just make your next sick ASCII art banner a bit easier to generate.

New Part Day: Flush-mount Touchscreen For Retro PC Build

December 11, 2023 by Donald Papp 19 Comments

I recently had the opportunity to purchase an early version of a new display, and it happened to be just the thing I needed to make a project work. That display is the Elecrow 11.6″ CrowVision touchscreen slated for release in 2024. Preorders are being accepted on Crowd Supply.

I had an idea for a retro-inspired PC build that was just waiting for a screen like this. I’ll talk about the display and what’s good about it, then showcase the build for which it was the missing piece. If you’ve got a project waiting for something similar, maybe this part will provide what you need or at least turn on some new ideas.

What Is It?

The CrowVision 11.6″ 1366 x 768 touchscreen has an HDMI input, USB output for touch data, and accepts 12 V DC. It’s made to interface easily with a Raspberry Pi or other SBC (single-board computer).

Personally I consider a display like this to be the minimum comfortable size for using desktop type applications in a windowed environment. Most displays in this space are smaller. But aside from that, what helps make it useful for embedding into a custom enclosure is the physical layout and design.

Since I was looking for the largest display that could be flush-mounted in an enclosure without a lot of extra space around the display’s sides, it was just what I needed. The integrated touchscreen is a nice bonus.

Continue reading “New Part Day: Flush-mount Touchscreen For Retro PC Build” →

Cowgol Development Environment Comes To Z80 And CP/M

December 10, 2023 by Donald Papp 7 Comments

Cowgol on Z80 running CP/M ties together everything needed to provide a Cowgol development environment (including C and assembler) on a Z80 running the CP/M operating system, making it easier to get up and running with a language aimed to be small, bootstrapped, and modern.

Cowgol is an experimental modern language for (very) small systems.

The Zilog Z80 was an 8-bit microprocessor common in embedded systems of the 1970s and 1980s, and CP/M was a contemporary mass-market operating system. As for Cowgol? It’s an Ada-inspired compiler toolchain and programming language aimed at very small systems, such as the Z80.

What’s different about Cowgol is that it is intended to be self-hosted on these small systems; Cowgol is written in itself, and is able to compile itself. Once one has compiled the compiler for a particular target architecture (for example, the Z80) one could then use that compiler on the target system to compile and run programs for itself.

Thankfully, there’s no need to start from scratch. The Cowgol on Z80 running CP/M repository (see the first link of this post) contains the pre-compiled binaries and guidance on using them.

Cowgol is still under development, but it works. It is a modern language well-suited to (very) small systems, and thanks to this project, getting it up and running on a Z80 running CP/M is about as easy as such things can get.

Thanks to [feinfinger] for the tip!

Intentional Filament Stringing Helps Santa Soar

December 7, 2023 by Donald Papp 3 Comments

Stringing is when a 3D printer’s hot end moves through open air and drags a wisp of melted plastic along with it. This is normally undesirable, but has in the past been done intentionally to create some unconventional prints. Moonlight Santa from [3dprintbunny] shows considerable refinement in the technique, complete with color changes that really make the result pop.

Using a 3D printer’s stringing in a constructive way is something that has been leveraged really well. We remember seeing a lion with a fantastic mane by combining this method with a little post-processing and a blast from a heat gun. The technique has also been applied to make brush bristles (the printer strings filament across two handles, and after printing it is cut in half to make two brushes.)

This isn’t [3dprintbunny]’s first rodeo, either. We loved seeing her show what kind of objects were possible by using clever design, with no reliance on custom G-code or weird slicer tricks. The color changes by filament swaps really make this new one stand out.

Making Visual Anagrams, With Help From Machine Learning

December 7, 2023 by Donald Papp 1 Comment

[Daniel Geng] and others have an interesting system of generating multi-view optical illusions, or visual anagrams. Such images have more than one “correct” view and visual interpretation.

What’s more, there are quite a few different methods on display: 90 degree flips and other (orthogonal) image rotations, color inversions, jigsaw permutations, and more. The project page has a generous number of examples, so go check them out!

The team’s method uses pre-trained diffusion models — more commonly known as the secret sauce inside image-generating AIs — to evaluate and work to combine the differences between different images, and try to combine and apply it in a way that results in the model generating a good visual result. While conceptually straightforward, this process wasn’t really something that could work without diffusion models driven by modern machine learning techniques.

The visual_anagrams GitHub repository has code and the research paper goes into details on implementation, limitations, and gives guidance on obtaining good results. Image generation is just one of the rapidly-evolving aspects of recent innovations, and it’s always interesting to see unusual applications like this one.

Mozilla Lets Folks Turn AI LLMs Into Single-File Executables

December 2, 2023 by Donald Papp 20 Comments

LLMs (Large Language Models) for local use are usually distributed as a set of weights in a multi-gigabyte file. These cannot be directly used on their own, which generally makes them harder to distribute and run compared to other software. A given model can also have undergone changes and tweaks, leading to different results if different versions are used.

To help with that, Mozilla’s innovation group have released llamafile, an open source method of turning a set of weights into a single binary that runs on six different OSes (macOS, Windows, Linux, FreeBSD, OpenBSD, and NetBSD) without needing to be installed. This makes it dramatically easier to distribute and run LLMs, as well as ensuring that a particular version of LLM remains consistent and reproducible, forever.

This wouldn’t be possible without the work of [Justine Tunney], creator of Cosmopolitan, a build-once-run-anywhere framework. The other main part is llama.cpp, and we’ve covered why it is such a big deal when it comes to running self-hosted LLMs.

There are some sample binaries available using the Mistral-7B, WizardCoder-Python-13B, and LLaVA 1.5 LLMs. Just keep in mind that if you’re on a Windows platform, only the LLaVA 1.5 will run, because it’s the only one that squeaks under the 4 GB limit on executable files that Windows has. If you run into issues, check out the gotchas list for troubleshooting tips.