π Smash The Mesh (STM) is a #GodotEngine add-on that lets you break 3D meshes into multiple pieces and apply physics to each fragment! Now in BETA, with caching for quick loading. Open-source and free to download! Get it here: https://t.co/9DiYo7RpDb π§ #Godot4#GameDev
@jam3scampbell A specific use case: I use a canvas document to manage my fridge contents, but when ChatGPT suggests recipes or ingredients, it often doesn't re-read the updated canvas and ends up recommending items that are no longer available.
@Saish_in3D@measure_plan@threejs I'm really glad you got it to build! I absolutely know how that feels, believe meπ
As for me, I wrestled with Bazel the dull way: PowerShell β Bazel CLI β error β fix β repeat... until I finally won.
@Saish_in3D@measure_plan@threejs Have you tried building a C API .so target in Bazel, even on Windows? The MediaPipe hand landmarker can compile into a DLL (or .so) that depends only on OpenCV. I had to tweak parts of the code for compatibility, but I used MSVC, not Clang. Could be a helpful starting point.
@_brunodeangelis@Hamster_Lite I haven't compiled with GPU support for portability reasons; performance wasn't critical for this demo.
To enable GPU with MediaPipe C/C++, you need the Bazel build system. I'd recommend starting on Linux: better support and simpler setup. My demo, instead, was built on Windows.
π Real-time AI hand tracking in #cplusplus and #raylib using Google #MediaPipe (C API build, no Python).
Double-hand detection Β· Smoothed landmarks Β· 20β25 ms CPU response.
Compiling MediaPipe for C/C++ wasn't trivial, but it works!
#handtracking#gamedev
@CarlosDelfinoCP Hi! Iβm considering sharing on GitHub after more testing. I'd also like to provide binaries for Windows/Linux/macOS (building it from source isnβt exactly straightforward). This setup uses MediaPipe Hands without Gesture Recognizer, so gestures need to be implemented separately.
@JoseLuAlba Thanks! Yes, I'm using 3D keypoints with linear interpolation and heuristics (like sudden wrist shifts or hand switches). I'm also exploring the 1β¬ filter (https://t.co/rQoQ1dDJ51). Gesture recognition isn't built-in yet, but I'm considering sharing after more testing.
@Hamster_Lite It's pretty impressive when you realize that this entire hand detection module, compiled as a C API, is under 8β―MB. The model itself is just around 7β―MB. Even cooler, the detection runs entirely on CPU. With GPU enabled, Iβd expect it to perform even better in a game engine βΊοΈ
Calling this playful #b3d test "Greasy-Ivy":
Sketching ivy in #CG with the #GreasePencil and growing it procedurally using #GeometryNodes and modifiers.
π First release of the #Rust content card feed system for my homepage is live! No more empty pages.
It's animated, responsive, and GPU-powered! π¦
Check it out: https://t.co/sYzq8Mhs8K
#creativecoding#indieweb
@HugoDuprez Thank you! After I finish this project, I should have learned a couple more things in Rust, and hopefully my homepage wonβt be empty anymore. π
As an experiment, I swapped the default macroquad::ui with #egui in my web demo. Interesting results:
WASM: 512 KB β 2.2 MB
Gzip: 215 KB β 1.1 MB
Brotli: 188 KB β 991 KB
Compression helps, but #egui adds ~1.5 MB!
#Rust#WASM#Brotli#Gzip