We present a learning-based method for synthesizing novel views of complex scenes using only unstructured collections of in-the-wild photographs. We build on Neural Radiance Fields (NeRF), which uses the weights of a multilayer perceptron to model th
e density and color of a scene as a function of 3D coordinates. While NeRF works well on images of static subjects captured under controlled settings, it is incapable of modeling many ubiquitous, real-world phenomena in uncontrolled images, such as variable illumination or transient occluders. We introduce a series of extensions to NeRF to address these issues, thereby enabling accurate reconstructions from unstructured image collections taken from the internet. We apply our system, dubbed NeRF-W, to internet photo collections of famous landmarks, and demonstrate temporally consistent novel view renderings that are significantly closer to photorealism than the prior state of the art.
For mobile robots navigating on sidewalks, it is essential to be able to safely cross street intersections. Most existing approaches rely on the recognition of the traffic light signal to make an informed crossing decision. Although these approaches
have been crucial enablers for urban navigation, the capabilities of robots employing such approaches are still limited to navigating only on streets containing signalized intersections. In this paper, we address this challenge and propose a multimodal convolutional neural network framework to predict the safety of a street intersection for crossing. Our architecture consists of two subnetworks; an interaction-aware trajectory estimation stream IA-TCNN, that predicts the future states of all observed traffic participants in the scene, and a traffic light recognition stream AtteNet. Our IA-TCNN utilizes dilated causal convolutions to model the behavior of the observable dynamic agents in the scene without explicitly assigning priorities to the interactions among them. While AtteNet utilizes Squeeze-Excitation blocks to learn a content-aware mechanism for selecting the relevant features from the data, thereby improving the noise robustness. Learned representations from the traffic light recognition stream are fused with the estimated trajectories from the motion prediction stream to learn the crossing decision. Furthermore, we extend our previously introduced Freiburg Street Crossing dataset with sequences captured at different types of intersections, demonstrating complex interactions among the traffic participants. Extensive experimental evaluations on public benchmark datasets and our proposed dataset demonstrate that our network achieves state-of-the-art performance for each of the subtasks, as well as for the crossing safety prediction.
Compared to LiDAR-based localization methods, which provide high accuracy but rely on expensive sensors, visual localization approaches only require a camera and thus are more cost-effective while their accuracy and reliability typically is inferior
to LiDAR-based methods. In this work, we propose a vision-based localization approach that learns from LiDAR-based localization methods by using their output as training data, thus combining a cheap, passive sensor with an accuracy that is on-par with LiDAR-based localization. The approach consists of two deep networks trained on visual odometry and topological localization, respectively, and a successive optimization to combine the predictions of these two networks. We evaluate the approach on a new challenging pedestrian-based dataset captured over the course of six months in varying weather conditions with a high degree of noise. The experiments demonstrate that the localization errors are up to 10 times smaller than with traditional vision-based localization methods.
