Deep-BrownConrady: Prediction of Camera Calibration and Distortion Parameters Using Deep Learning and Synthetic Data

2025-01-24

AI Summary

Conventional camera calibration relies on multi-view image sequences, limiting applicability in scenarios where only a single image is available. Method: This paper proposes the first end-to-end deep learning method for single-image camera calibration, directly regressing all intrinsic parameters of the Brown–Conrady model—focal length, principal point, and radial/tangential distortion coefficients—from a single input image. We design a ResNet-based regression architecture and introduce the first high-fidelity synthetic dataset covering the full parameter space, generated using AILiveSim. Training employs a real-synthetic hybrid strategy to bridge domain gaps. Contribution/Results: On real-world images, our method achieves an average reprojection error < 0.35 pixels, principal point offset error < 1.2 pixels, and relative distortion coefficient error < 8.7%—significantly outperforming existing calibration-free methods. This work establishes, for the first time, the feasibility and practicality of single-image deep learning–based calibration, offering a lightweight, robust calibration paradigm for autonomous driving, robotics, and VR applications.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

This research addresses the challenge of camera calibration and distortion parameter prediction from a single image using deep learning models. The main contributions of this work are: (1) demonstrating that a deep learning model, trained on a mix of real and synthetic images, can accurately predict camera and lens parameters from a single image, and (2) developing a comprehensive synthetic dataset using the AILiveSim simulation platform. This dataset includes variations in focal length and lens distortion parameters, providing a robust foundation for model training and testing. The training process predominantly relied on these synthetic images, complemented by a small subset of real images, to explore how well models trained on synthetic data can perform calibration tasks on real-world images. Traditional calibration methods require multiple images of a calibration object from various orientations, which is often not feasible due to the lack of such images in publicly available datasets. A deep learning network based on the ResNet architecture was trained on this synthetic dataset to predict camera calibration parameters following the Brown-Conrady lens model. The ResNet architecture, adapted for regression tasks, is capable of predicting continuous values essential for accurate camera calibration in applications such as autonomous driving, robotics, and augmented reality. Keywords: Camera calibration, distortion, synthetic data, deep learning, residual networks (ResNet), AILiveSim, horizontal field-of-view, principal point, Brown-Conrady Model.

Problem

• Machine Learning
• Camera Calibration
• Single Image Prediction

Innovation

• Advanced Machine Learning
• ResNet Model Adaptation
• AILiveSim Tool