The next chapter about transformers is up on YouTube, digging into the attention mechanism: youtu.be/eMlx5fFNoYc The model works with vectors representing tokens (think words), and this is the mechanism that allows those vectors to take in meaning from context.

Diffusion models turn the data into a mixture of isotropic Gaussians, and so struggle to capture the underlying structure when trained on small datasets. In our new #ECCV2024 paper, we introduce RS-IMLE, a generative model that gets around this issue. Website:

MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion TL;DR: DUSt3R keeps impressing. Contributions (cited): • We introduce Motion DUSt3R (MonST3R), a geometry-first approach to dynamic scenes that directly estimates geometry in the form of pointmaps,

An Overview of BRDF Models digibug.ugr.es/bitstream/hand…

You can now preorder our new Vulkan rendering cookbook! lnkd.in/d4TxbMK2 As always, it takes you from a simple triangle to creating your own glTF2 renderer, along with plenty of other exciting recipes. The book uses LightweightVK and explores its implementation.

Reinventing the Geometry Pipeline: Mesh Shaders in DirectX 12 | Shawn Ha... youtu.be/CFXKTXtil34?si… via YouTube

Here are my experiments with traversal methods so far. github.com/Niravaana/HIP-… Implemented for 2 - BVH as of now, will extend it to wide bvhs. Methods implemented so far 1. if/ if traversal 2. while/while traversal 3. speculative while/while 4. Restart trail

How does this video help translate to a 15-40X file size compression for Radiance Field method, Gaussian Splatting? Tune in this Thursday as we speak to Wieland Morgenstern about his paper, Compact 3D Scene Representation via Self-Organizing Gaussian Grids. Link: youtube.com/@ViewDependent

The simulated annealing algorithm explained with an analogy to a toy youtu.be/eBmU1ONJ-os?si… via YouTube

SCube: Instant Large-Scale Scene Reconstruction using VoxSplats Abstract: Our method encodes reconstructed scenes using a novel representation VoxSplat, which is a set of 3D Gaussians supported on a high-resolution sparse-voxel scaffold. To reconstruct a VoxSplats from images,

Sorting for Rendering linebender.org/wiki/gpu/sorti…

SplatDoF: Real-Time Depth of Field for Gaussian Splatting kth.diva-portal.org/smash/record.j…

3DGS.zip: A survey on 3D Gaussian Splatting Compression Methods w-m.github.io/3dgs-compressi…

Denoising in Cycles: Past, present and future — Blender Conference 2024 youtu.be/jirpTs8DBPY?si… via YouTube

Ok, instead of just bitching on X for fun, I started collecting some of the general GPU programming improvement suggestions here: docs.google.com/document/d/1My… (link provides comment access too). I'll try to collect ideas in that doc in the future too.

2014 LOW LEVEL OPTIMIZATIONS FOR GCN michaldrobot.com/wp-content/upl…

Paper: arxiv.org/abs/2411.14974 Project: convexsplatting.com

2013 Level of Detail for Real-Time Volumetric Terrain Rendering animation.rwth-aachen.de/media/papers/2…

Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes nv-tlabs.github.io/nglod/assets/n…