InFlux++: Real and Synthetic Data for Estimating Dynamic Camera Intrinsics

Summary

InFlux++ is a dataset and benchmark suite for dynamic camera intrinsics estimation, addressing the common assumption in 3D vision algorithms that camera intrinsics remain fixed throughout a video. The release consists of two components. InFlux++ Synth is a large-scale procedurally generated synthetic video dataset with over 441,000 annotated frames and 1,841 high-resolution videos, providing accurate per-frame ground-truth intrinsics. InFlux++ Real extends a real-world benchmark with more than 514,000 new frames and 334 high-resolution videos. The authors, from Princeton (Erich Liang, Caleb Kha-Uong, Chinmaya Saran, Sreemanti Dey, David W. Liu, Junhan Ouyang, Benjamin Zhou, Jia Deng), show that fine-tuning existing intrinsics prediction methods on InFlux++ Synth improves focal length estimation accuracy on both real-world benchmarks, indicating that synthetic supervision is a promising direction for RGB-based camera intrinsics prediction. The paper is available on arXiv as 2607.05389, with a project website at https://influx.cs.princeton.edu/.