Overview
Visualskies’ virtual capture of the famous Le Train Bleu restaurant in Paris was completed using high-end photogrammetry and laser scanning. With this data we created a high resolution 3D reconstruction of the ornate dining salon. Our focus was on optimising this accurate scanned asset for photo-realistic real time visualisation which is increasingly requested by our clients for Virtual Production and a wide range of AR/VR applications.
We cleaned the geometry and then split it up into its component objects which were then processed separately. After detailed retopology, uv mapping, texture reconstruction and texture cleanup, the objects were precisely combined. Next, shading, lighting and photo-real visualisations of the salon in the Unreal Engine were created and, to complement these, we generated camera animations rendered as a video sequence. For processing the data Reality Capture, ZBrush, Maya, Mudbox, Meshmixer and Unreal Engine were extensively utilised.
Project
Le Train Bleu
Client
N/A
What We Did
3D Scanning and Data Processing
See the final animation below
The data was captured over one night using LiDAR and photogrammetry.
The team
Data Capture
LiDAR data
Data Processing
We processed the precise 3D laser scanner data in Leica Cyclone and generated a dense point cloud upon which the detailed geometry is based.
High resolution geometry
Using cutting edge 3D scanning techniques we generated a highly detailed 3D reconstruction of the environment.
High resolution textures
With photogrammetry we reconstructed high quality textures which support the photo-realistic look.
Real time visualisation
After data processing we visualised the 3D scan in real time in the Unreal Engine.
The Tools We Deployed
‘Ground Truth’
Terrestrial LiDAR
Utilising the high accuracy of the Leica P40 we were able to capture all the subtle details in high fidelity. The precise point cloud collected by the scanner allowed us to combine photogrammetry with the LiDAR data to create a highly detailed and photo-realistic 3D model.
Textured Dataset
Photogrammetry
We captured over 2000 images of Le Train Bleu and combined them with the LiDAR data.