Wide Field-of-View 3D Reconstruction via Camera Model Conditioning

Dongki Jung1 Jaehoon Choi1 Adil Qureshi1 Somi Jeong2
Dinesh Manocha1 Suyong Yeon2
1University of Maryland, College Park 2NAVER LABS
💻 GitHub 📄 Paper

Main Contribution

We present Wid3R, a feed-forward neural network for visual geometry reconstruction that supports wide field-of-view camera models.

Model Pipeline

Training Dataset

Training Dataset

Network Architecture

Network Architecture

Interactive Demo

We visualize some in-the-wild videos, demonstrating the generalizability of our method.

💡 Please Note: To ensure a smooth interactive experience, the point clouds in this demo are downsampled. The quality of the full model is much higher.

Rotate

Hold & Drag Left Mouse

Pan

Hold & Drag Right Mouse

Zoom

Use Mouse Scroll Wheel