7th Datathon

Title: 7th Annual ASCE VIMS Datathon Competition 2026

Scope: The 2026 Datathon focuses on applying 3D Gaussian Splatting (3DGS) to reconstruct high-fidelity 3D models of a selected indoor built environment from multimodal visual data. Unlike traditional mesh-based modeling, 3DGS enables photorealistic, efficient, and scalable scene reconstruction from image or video streams, making it highly relevant for digital twins, robotics simulation, facility management, and immersive visualization.

Participants will reconstruct a physically and visually coherent 3D model of the designated site using 3DGS or comparable neural rendering techniques. The objective is to produce a scene representation that is not only visually compelling but also usable for downstream applications such as robotic navigation, inspection simulation, safety analysis, or digital twin integration.

Organizer: ASCE VIMS Committee

Time: Feb -June 2026

Dataset: A comprehensive dataset of a selected indoor environment will be distributed to all registered teams. The dataset will include:

• Lidar point clouds for geometry comparison
• High-resolution image and/or video sequences for 3DGS reconstruction
• Camera poses and calibration information (in Colmap convention)

Winners:

First place:
 

Runner-up:

Third place: