Procedural Map Creation from Sensor Data | Paul Ambrosiussen / Joey Kowalewicz | EPC 2025 HIVE
In this session, Paul and Joey from Applied Intuition will dive into two production workflows for generating procedural maps from multimodal sensor data, including lidar, camera imagery, and auxiliary metadata. The talk will cover a traditional pipeline centered around mesh and heightfield generation, as well as an alternative pipeline leveraging Gaussian Splatting for efficient scene reconstruction. The talk covering Houdini and Unreal Engine workflows will highlight key technical challenges, data conditioning steps, and tradeoffs in fidelity, performance, and scalability.
Paul Ambrosiussen is a freelance Senior Technical Artist at Bismuth Consultancy, specializing in procedural content generation, pipeline development, and the use of machine learning in creative production. He works extensively with tools like Houdini and Unreal Engine to build scalable solutions for studios. Previously, he led tool and workflow development at SideFX Labs, collaborating with teams across games, film, and tech. Paul thrives on solving complex pipeline challenges, prototyping experimental workflows, and bridging the gap between artistry and technology through talks, workshops, and community-driven initiatives.
Joey Kowalewicz is a staff Houdini TA at Applied Intuition with a passion for data processing and content generation for offroad simulation. He primarily works with lidar and other forms of real world sensor data to generate synthetic data sets that mimic reality. Previous, Joey has worked as an FX TD at Liaka, Blue Sky and Pixar Studios amoung other various commercial and episodic studios with a focus on fluid dynamics. Processing and manipulating the large scale fluid simulations has given him a respect for data analytics and precision accuracy leading to where he is today.
00:00:00 Introduction to procedural map creation
00:03:38 Overview of segmented workflow
00:07:05 Handling outlier elevation data on roads
00:10:20 Map generation from camera imagery
00:13:02 Capturing 360-degree scene views
00:15:42 Using extrinsics for 3D scene reconstruction
00:18:28 Projecting and remapping depth maps
00:21:10 Explanation of process and methodology
00:23:50 Rendering point clouds using discs
00:26:47 Wrap-up and final notes
See more talks from EPC 2025:
• EPC 2025 HIVE