Published:
A new architecture leveraging fixed pretrained semantic segmentation networks to guide self-supervised representation learning via pixel-adaptive convolutions.
Published:
A new dataset (STIP) and graph neural network operating on scene graphs to predict pedestrian crossing intent.
Published:
A new state of the art in self-supervised monocular depth estimation (PackNet), a new benchmark (DDAD), and weak velocity supervision.
Published:
An automatic annotation pipeline to recover 9D cuboids and 3D shapes from pre-trained off-the-shelf 2D detectors and sparse LIDAR data.
Published:
A novel panoptic segmentation method featuring parameter-free instance mask reconstruction, state-of-the-art accuracy, and real-time inference.
Published:
A model that learns to distill spatio-temporal object interactions for video captioning.
Published:
A review of the literature and current state of differentiable rendering, its applications, and open research problems.
Published:
A hierarchical reinforcement and imitation learning (H-ReIL) approach for learning to drive in near-accident scenarios.
Published:
Neural Ray Surfaces (NRS) are convolutional networks that represent pixel-wise projection rays, approximating a wide range of cameras. NRS are fully differentiable and can be learned end-to-end from unlabeled raw videos.
Published:
Novel self-supervised method for textured 3D shape reconstruction and pose estimation of rigid objects with the help of strong shape priors and 2D instance masks.
Published:
Predicted Endpoint Conditioned Network (PECNet) for flexible human trajectory prediction via endpoint inference.
Published:
A method to automatically find planner-specific defects of autonomous vehicles in simulation.
Published:
Reinforcement Learning of autonomous driving agent policies to balance driving skills and behavioral diversity.
Published:
A novel representation of perceptual uncertainty for learning to plan via behaviorial cloning.
Published:
An end-to-end deep learning framework for LIDAR-based flow estimation in 2.5D bird’s eye view (BeV). We show it boosts tracking performance on a real-world autonomous car.
Published:
A risk-sensitive game-theoretic planning algorithm to model complex multi-agent interactions yielding more time-efficient, intuitive, and safe behaviors when facing underlying risks and uncertainty.
Published:
An online framework for safe and efficient crowd-robot interaction using sampling based trajectory forecasting with risk-sensitive optimal control.
Published:
We propose a data-dependent regularization technique for heteroskedastic and imbalanced datasets.
Published:
We show monocular depth estimation methods can work inside small fluid-filled visuotactile sensors like Soft-Bubbles.
Published:
Published:
We learn per pixel feature vectors that simultaneously encode instance- and category-level discriminative information.
Published:
Learning a single network for dialable depth perception, with or without sparse LiDAR input.