FIORD: A Fisheye Indoor-Outdoor Dataset with LIDAR Ground Truth for 3D Scene Reconstruction and Benchmarking

2025-04-02

AI Summary

Existing 3D scene reconstruction methods rely on narrow-field-of-view perspective images, requiring large image collections and complex Structure-from-Motion (SfM), thus suffering from poor scalability to large-scale scenes. This work introduces the first omnidirectional fisheye image dataset specifically designed for 3D reconstruction, covering 10 real-world indoor and outdoor scenes. It uniquely integrates 200° fisheye imaging with high-accuracy LiDAR-derived dense ground truth, and provides synchronized image sequences, calibrated intrinsic/extrinsic parameters, SfM-sparse point clouds, and co-registered dense point clouds. The dataset effectively alleviates key bottlenecks—including inefficient large-area coverage, severe occlusion and reflective artifacts, and unreliable geometric evaluation—enabling robust and fair benchmarking of novel view synthesis and reconstruction methods. Baseline evaluations using Gaussian Splatting and Nerfacto demonstrate its superiority in both reconstruction efficiency and geometric accuracy assessment.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications

Abstract

The development of large-scale 3D scene reconstruction and novel view synthesis methods mostly rely on datasets comprising perspective images with narrow fields of view (FoV). While effective for small-scale scenes, these datasets require large image sets and extensive structure-from-motion (SfM) processing, limiting scalability. To address this, we introduce a fisheye image dataset tailored for scene reconstruction tasks. Using dual 200-degree fisheye lenses, our dataset provides full 360-degree coverage of 5 indoor and 5 outdoor scenes. Each scene has sparse SfM point clouds and precise LIDAR-derived dense point clouds that can be used as geometric ground-truth, enabling robust benchmarking under challenging conditions such as occlusions and reflections. While the baseline experiments focus on vanilla Gaussian Splatting and NeRF based Nerfacto methods, the dataset supports diverse approaches for scene reconstruction, novel view synthesis, and image-based rendering.

Problem

• Lack of fisheye datasets for 3D scene reconstruction
• Scalability issues with narrow FoV image datasets
• Need for geometric ground-truth in challenging conditions

Innovation

• Fisheye image dataset for 360-degree coverage
• LIDAR-derived dense point clouds as ground-truth
• Supports Gaussian Splatting and NeRF methods