Short-Term Prediction and Multi-Camera Fusion on Semantic Grids
This work addresses the need for improved environment representations in autonomous systems, though it appears incremental as it builds on existing semantic segmentation and fusion techniques.
The authors tackled the problem of generating a predictive, semantically-interpretable environment representation for autonomous systems by fusing multiple camera sequences and making short-term predictions, showing that their architecture significantly outperforms a model-driven baseline in a category-based evaluation on the Cityscapes dataset.
An environment representation (ER) is a substantial part of every autonomous system. It introduces a common interface between perception and other system components, such as decision making, and allows downstream algorithms to deal with abstracted data without knowledge of the used sensor. In this work, we propose and evaluate a novel architecture that generates an egocentric, grid-based, predictive, and semantically-interpretable ER. In particular, we provide a proof of concept for the spatio-temporal fusion of multiple camera sequences and short-term prediction in such an ER. Our design utilizes a strong semantic segmentation network together with depth and egomotion estimates to first extract semantic information from multiple camera streams and then transform these separately into egocentric temporally-aligned bird's-eye view grids. A deep encoder-decoder network is trained to fuse a stack of these grids into a unified semantic grid representation and to predict the dynamics of its surrounding. We evaluate this representation on real-world sequences of the Cityscapes dataset and show that our architecture can make accurate predictions in complex sensor fusion scenarios and significantly outperforms a model-driven baseline in a category-based evaluation.