CULTURE3D: Cultural Landmarks and Terrain Dataset for 3D Applications

This repository contains a collection of large-scale, fine-grained 3D datasets using high-resolution images captured from various locations worldwide. These datasets support multiple applications, including reconstruction, localization, semantic segmentation, and virtual reality.

Open Source Data & Project Updates

Our project is fully open-sourced, and all dataset files are available on our Google Drive. In addition, we are actively working on our Unreal Engine project, which will be uploaded as soon as it’s completed. Below is a quick overview of our current resources and their statuses:

Google Drive Contents Google Drive Link

• Raw Images
  High-resolution JPEG images for each capture.
• Point Cloud Data
  Dense and sparse point clouds in .ply and .pcd formats.
• Processed Data
  Includes COLMAP outputs, RealityCapture projects, and TLS-based ground truth files.
• Virtual Reality Files
  VR-ready 3D scenes with high-detail meshes and point clouds using realisy capture.
• Unreal Engine Project
  Currently pending upload – to be finished and shared later. (uploading, pending next week)

For an in-depth look at our methodology and results, please refer to our Arxiv Paper (to be updated). You can also explore the complete open-source dataset and supporting code on our GitHub Repository.

Datasets Overview

We provide datasets for iconic landmarks and cultural heritage sites, each meticulously captured to ensure high accuracy and detail. Below are the available datasets with access links:

1. Buckingham Palace Dataset
   
   A detailed reconstruction of Buckingham Palace, emphasizing architectural elements like gates and sculptural decorations.

2. Cambridge Campus Dataset
   
   High-resolution modeling of Cambridge University, ideal for urban planning and mixed-reality applications.

   This specific dataset contains a Virtual Reality (VR) scene which is available for download and view in future.

3. Egyptian Pyramids and Sphinx Dataset
   
   Captures intricate carvings and erosion patterns of these ancient structures.

4. Louvre Museum Dataset
   
   Comprehensive exterior depiction of the Louvre Museum with ornate architectural details.

5. Leaning Tower of Pisa Dataset
   
   Detailed reconstruction focusing on the unique tilt and masonry of this historic structure.

   This specific dataset contains a Virtual Reality (VR) scene which is available for download and view in future.

6. Stonehenge Dataset
   
   Captures the iconic rocks with emphasis on spatial arrangement and structural details.

7. Petra Dataset
   
   Offers intricate details of natural stone formations and surrounding environments.

   This specific dataset contains a Virtual Reality (VR) scene which is available for download and view in future.

8. Trafalgar Square Dataset
   
   High-resolution models of statues and surrounding structures within Trafalgar Square.

9. National Art Gallery Dataset
   
   
   Captures gallery interiors, including paintings, sculptures, and architectural features.

   This specific dataset contains a Virtual Reality (VR) scene which is available for download and view in future.

10. Forbidden City Dataset
    
    Detailed depiction of Chinese architectural heritage with intricate carvings and roof patterns.

11. Longmen Grottoes Dataset
    
    High-resolution carvings of the Longmen Grottoes, capturing natural weathering effects.

    This specific dataset contains a Virtual Reality (VR) scene which is available for download and view in future.

---

Dataset Format

Raw Data

• Images: High-resolution JPEG images named as <image_number>.jpg.
• Point Cloud Data: Dense and sparse point clouds in .ply and .pcd formats.

Processed Data

• COLMAP Outputs: Includes reconstruction data in the COLMAP format.
• RealityCapture Projects: Reconstruction results with features such as camera parameters, dense and sparse point clouds, undistorted images, and depth maps.
• Ground Truth: TLS (Terrestrial Laser Scanning)-based high-accuracy reconstruction files.

Virtual Reality Files

Reconstructed VR-ready 3D scenes with high-detail meshes and point clouds.

---

Applications

• Reconstruction and Localization: Benchmarks for SLAM, SfM, and radiance field-based methods like Neural Radiance Fields and Gaussian Splatting.
• Virtual Reality Integration: Ready-to-use models for VR applications with high-resolution details.
• Cultural Preservation: Enables digital preservation and restoration planning.
• Research and Training: Supports training machine learning models for segmentation, reconstruction, and navigation.