Scalable Uncertainty for Computer Vision with Functional Variational Inference
This work addresses the need for safe and reliable deployment of deep learning in real-world computer vision applications by providing scalable uncertainty quantification, though it is incremental as it builds on existing variational inference and Gaussian Process methods.
The paper tackles the problem of quantifying uncertainty in deep learning for computer vision by using functional variational inference with Gaussian Processes, achieving predictive uncertainty estimates with a single forward pass and enabling fast training for high-dimensional tasks like depth estimation and semantic segmentation.
As Deep Learning continues to yield successful applications in Computer Vision, the ability to quantify all forms of uncertainty is a paramount requirement for its safe and reliable deployment in the real-world. In this work, we leverage the formulation of variational inference in function space, where we associate Gaussian Processes (GPs) to both Bayesian CNN priors and variational family. Since GPs are fully determined by their mean and covariance functions, we are able to obtain predictive uncertainty estimates at the cost of a single forward pass through any chosen CNN architecture and for any supervised learning task. By leveraging the structure of the induced covariance matrices, we propose numerically efficient algorithms which enable fast training in the context of high-dimensional tasks such as depth estimation and semantic segmentation. Additionally, we provide sufficient conditions for constructing regression loss functions whose probabilistic counterparts are compatible with aleatoric uncertainty quantification.