AndreasR

@ibaiGorordo @rerundotio Looks awesome! :) I'm a dev at @rerundotio - curious why couldn't you use Rerun's built-in 3d line strip for this? Was the look not quite right or something else that didn't pan out?

Check out our model in action! Our model’s accuracy feels even more impressive when visualized with @rerundotio (It would be nice to have flat 3D lines, without using mesh🙏)

Tried out the new video support in @rerundotio 0.19 by visualizing @Apple's DepthPro model with @Gradio on @huggingface Spaces. Super cool to see it in action! oxkitsune-rerun-ml-depth-pro.hf.space

It's happening! Rerun is now not just a visualizer but also a multi-model database you can query 👇

This round of MacOS's Finder vs @rerundotio goes toooo Rerun 😁 Oof, video is hard! 😅

@AiFlux @rerundotio Also btw. the tweet is about the native viewer. On the web we're using the WebCodecs APIs for all video decoding

@AiFlux @rerundotio if you use our Python/Rust/C++ apis to add text to the data it's fairly easy, but there's no way to do that in Javascript. Rerun renders to a WebGL/WebGPU canvas that you'd need to overlay then

At first, we thought software decoding AV1 might be too slow for our purposes. But after we pushed the planar YUV (output of the decoder) --> RGB (input of our renderer) conversion to the GPU it's pretty darn fast! Playback at 20x speed for this video, and not a single stutter!

Build it open source and make it run everywhere, this is what you get 👇

Video support for the upcoming 0.19 release is coming along nicely 🎥 👀 For this release it will be limited to the web viewer, only mp4 containers (no streaming yet), and supporting AV1 and H.264. But we're just getting started. Stay tuned!

@ernerfeldt wait what iOS introduced that only now? Having tri-lingual keyboard for ages on my phone ;-)

I can't believe we're done! Time Ghost A short film we on the Demo Team have been working on for a while youtu.be/o1JIK5W3DRU Like all our team’s demos, it runs in real-time at 30fps in Unity. Over the next days I’ll be posting rendering and tech art breakdowns in this 🧵

Some serious image format testing, using ethically sourced Lena picture mortenhannemose.github.io/lena/ by @MortenHannemose: github.com/rerun-io/rerun…

@lucasmeijer @MrChickenRocket see pee pee. thanks! Already had a hard time taking C++ all that seriously these days. That could be the coffin nail 😁

@MrChickenRocket We all say eye Elle two see pee pee

seeing openai's slomo carcrash of model names, i feel somewhat less bad about some of the horrible names we picked for unity features. (hello il2cpp)

Rerun 0.18: Columnar API and new data architecture for faster writes & ingestion + lower memory use. This release brings orders of magnitude faster writes and ingestion times and lower memory usage for many long time-series datasets. It’s powered by a new column chunk based architecture. The new rr.send_columns APIs makes it easier and faster to send pre-recorded data to Rerun without having to log it. For good measure, we also threw in mesh and 3D shape instancing through the InstancePose archetype as well as a new shape primitive for 3D ellipsoids. 0.18 is another big step forward on our quest to help robotics and embodied AI teams unify their visualization and data stacks by making Rerun fly through even more kinds of datasets!

The upcoming Rerun 0.17 brings a huge increase in explicit control over your visualizations 🚀 In this video for example we dynamically change the visualizer that is applied on an entity from Mesh3D to Points3D and then edit the radius of the points. Because this mesh has vertex colors, the point cloud automatically becomes colored as well. All of this can be done either from UI or from code. We’ll release 0.17 next week. Stay tuned.

@ernerfeldt Someone should really make cargo use yaml instead of toml. Cargo.yaml. Great idea! 😛 /s

HOT3D in Rerun! 🫱⌨️🫲 HOT3D is a new benchmark dataset for vision-based understanding of 3D hand-object interactions by @TweetEdMiller and the Project Aria team at @AIatMeta.

@ernerfeldt @ExpertOfNil Admittedly no-default-features doesn't make much sense for wgpu - you can't render anything then. 11.57s for default features Naga becomes heavier quite quickly

@ernerfeldt @ExpertOfNil Probably got fixed in 0.20: Just tried `cargo clean && cargo build --timings -p wgpu --no-default-features`, takes 9.7 on my macbook. Interesting how bad num-traits is Oo.