In Depth Bayesian Semantic Scene Completion
This work addresses the challenge of predicting occluded areas in 3D scenes for applications like robotics or autonomous systems, but it is incremental as it builds on existing methods by adding Bayesian uncertainty estimation.
The paper tackles the problem of Semantic Scene Completion by constructing a Bayesian Convolutional Neural Network (BCNN) to predict 3D semantic segmentation and model uncertainty, showing that it performs equal or better to standard CNNs on MNIST with unseen digits and outperforms them on the SUNCG dataset with a new category at test time, achieving better Intersection over Union, Average Precision, and separation scores.
This work studies Semantic Scene Completion which aims to predict a 3D semantic segmentation of our surroundings, even though some areas are occluded. For this we construct a Bayesian Convolutional Neural Network (BCNN), which is not only able to perform the segmentation, but also predict model uncertainty. This is an important feature not present in standard CNNs. We show on the MNIST dataset that the Bayesian approach performs equal or better to the standard CNN when processing digits unseen in the training phase when looking at accuracy, precision and recall. With the added benefit of having better calibrated scores and the ability to express model uncertainty. We then show results for the Semantic Scene Completion task where a category is introduced at test time on the SUNCG dataset. In this more complex task the Bayesian approach outperforms the standard CNN. Showing better Intersection over Union score and excels in Average Precision and separation scores.