DeepBrownConrady: Predicting Camera Calibration and Distortion Parameters Using Deep Learning and Synthetic Data
Date:
I presented my paper DeepBrownConrady (DBC) at AI Day 2025, organized by the Finnish Center for Artificial Intelligence (FCAI). Click on the heading for more details.
The talk showcased how DeepBrownConrady leverages deep learning trained on large-scale synthetic data to directly predict camera calibration and lens distortion parameters, providing a scalable and simulation-driven alternative to traditional calibration pipelines for vision-based systems.
I discussed:
Motivation for learning calibration parameters directly from images
The role of high-fidelity synthetic data in supervision
Model design and prediction of Brown–Conrady distortion parameters
Generalization to real-world datasets and downstream vision tasks
The presentation was part of AI Day 2025 in Helsinki and sparked valuable discussions around synthetic data, camera geometry, and robustness in vision systems.
