Efficient Global Optimization of Non-differentiable, Symmetric Objectives for Multi Camera Placement
This addresses camera placement optimization for computer vision practitioners, presenting an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of optimally placing and orienting multiple cameras in 3D scenes for applications like 3D reconstruction, surveillance, and pedestrian tracking, proposing a globally convergent, parallelizable algorithm that handles non-differentiable objectives efficiently.
We propose a novel iterative method for optimally placing and orienting multiple cameras in a 3D scene. Sample applications include improving the accuracy of 3D reconstruction, maximizing the covered area for surveillance, or improving the coverage in multi-viewpoint pedestrian tracking. Our algorithm is based on a block-coordinate ascent combined with a surrogate function and an exclusion area technique. This allows to flexibly handle difficult objective functions that are often expensive and quantized or non-differentiable. The solver is globally convergent and easily parallelizable. We show how to accelerate the optimization by exploiting special properties of the objective function, such as symmetry. Additionally, we discuss the trade-off between non-optimal stationary points and the cost reduction when optimizing the viewpoints consecutively.