Source Themes

f -Cal: Calibrated aleatoric uncertainty estimation from neural networks for robot perception

We present a simple approach that uses a variational loss to enforce calibration in probabilistic regression networks.

Taskography: Evaluating robot task planning over large 3D scene graphs

We present a large-scale benchmark and performant approaches for long-horizon task planning over large 3D scene graphs

DRACO: Weakly supervised dense reconstruction and canonicalization of objects

We present a weakly supervised approach that reconstructs objects in a canonical coordinate space.

gradSim: Differentiable simulation for system identification and visuomotor control

Differentiable models of time-varying dynamics and image formation pipelines result in highly accurate physical parameter estimation from video

AutoLay: Benchmarking Monocular Layout Estimation

We present a dataset and introduce a new benchmark for *amodal* layout estimation from monocular imagery.

Probabilistic Object Detection: Strengths, Weaknesses, Opportunities

We present a survey of current probabilistic object detection techniques, and identify promising avenues for further research.

gradSLAM: Dense SLAM meets automatic differentiation

We present end-to-end differentiable dense SLAM systems that open up new possibilites for integrating deep learning and SLAM.

MapLite: Autonomous intersection navigation without detailed prior maps

MapLite is a one-click autonomous navigation system for a vehicle that only uses OpenStreetMap data and local sensing (**Best paper award, RAL 2019**).

MonoLayout: Amodal scene layout from a single image

We present a neural network that "hallucinates" the layout of a road scene from a single image, including scene parts that are outside the bounds of the image.

Multi-object monocular SLAM for dynamic environments

We present a monocular object SLAM system that tracks not just the camera, but also other moving objects in the scene.