Skip to content
Drix10 Blog

4D Gaussians rendering 185k points in 0.4ms on 3090 Ti

, 5 items in Spatial Computing, 3 min read

In this digest (5 items)

4D Gaussians rendering 185k points in 0.4ms on a 3090 Ti using Rust/webgl. It shares the same WebGL context as threejs, so depth tests and occlusion work automatically.

Key points

  • Rendering speed: 0.4 ms on a 3090 Ti for 185k 4D Gaussians

  • Built in Rust/webgl, sharing context with threejs for depth tests and occlusion

Sources

PlantCraft Open-Source Software for Plant Management and Analysis Launch

PlantCraft is open-source software for plant management and analysis that integrates data history, Python analysis, alarms, and reports. It adds natural-language AI queries and supports offline, equipment-side read-only access.

Key points

  • Open Source (Apache-2.0) license

  • Developer: Jônathan Dambros

Sources

Spark Radix experimental release – Gaussian splat renderer on three.js

Spark Radix experimental release adds a Gaussian splat renderer built on three.js. It enables rendering and exploring detailed 3D worlds directly in the browser. The release is available today for developers to try and test.

Key points

  • Release: experimental version of Spark Radix announced

  • Feature: Gaussian splat renderer built on three.js for browser rendering

Sources

AI Analyzes Basketball Footage Provides Audio Commentary

AI Analyzes Basketball Footage provides audio commentary on game flow using YOLO detection and ByteTrack tracking. It detects players, ball, and rim, tracks positions and speeds, and operates in post-processing mode.

Key points

  • Detection: Detects players, ball, and rim using YOLO

  • Tracking: Tracks players with ByteTrack to estimate position, speed, and defensive spacing

  • Processing: Currently post-processing, not real-time processing

Sources

Boxing Coach Without a Camera Tracks Arms in 3D with IMU

Boxing Coach Without a Camera tracks arms in 3D using an IMU wristwatch and smartphone. It provides audio guidance for 10 punch types and 6 combo types and scores form, speed, and estimated power.

Key points

  • Measurement: wristwatch + smartphone on upper arm tracks movements in 3D.

  • Audio guidance: supports 10 punch types and 6 combo types.

  • Scoring: evaluates form, speed, and estimated power; power estimate based on physique and requires an Android-compatible smartwatch.

Sources