3D Object Detection

DVF: Dense Voxel Fusion for 3D Object Detection

Anas Mahmoud, and Steven L. Waslander
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023

we propose Dense Voxel Fusion (DVF), a sequential fusion method, that first assigns voxel centers to the 3D location of the occupied LiDAR voxel features.

PDV: Point Density-Aware Voxels for LiDAR 3D Object Detection

Jordan S.K. Hu, Tianshu Kuai and Steven L. Waslander

Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2022

We propose a Point Density-Aware Voxel network (PDV) to resolve LiDAR’s diverging point pattern issues by leveraging voxel point centroid localization and feature encodings. ​

Bayesian Embeddings for Few-Shot Open World Recognition

John Willes, James Harrison, Ali Harakeh, Chelsea Finn, Marco Pavone, and Steven Waslander
IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2022 ​

We combine Bayesian non-parametric class priors with an embedding-based scheme to yield a highly flexible framework which we refer to as few-shot learning for open world recognition (FLOWR).. ​

Pattern-Aware Data Augmentation for LiDAR 3D Object Detection

Jordan S.K. Hu and Steven L. Waslander
IEEE International Intelligent Transportation Systems Conference (ITSC), 2021 ​

Data augmentation technique that uses the diverging LiDAR point pattern to simulate objects at farther distances. ​

UrbanNet: Leveraging Urban Maps for Long Range 3D Object Detection

Juan Carrillo, and Steven L. Waslander
IEEE International Intelligent Transportation Systems Conference (ITSC), 2021 ​

We present UrbanNet, a modular architecture for long range monocular 3D object detection with static cameras. ​

Categorical Depth Distribution Network for Monocular 3D Object Detection

Cody Reading, Ali Harakeh, Julia Chae, Steven L. Waslander

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR, Oral), 2021

​

Monocular 3D object detection pipeline that estimates pixel-wise categorical depth distributions to accurately locate image information in 3D space. 

CG Stereo: Confidence Guided Stereo 3D Object Detection with Split Depth Estimation

Chengyao Li, Jason Ku, and Steven L. Waslander

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020

Confidence-guided stereo 3D object detection pipeline that uses separate decoders for foreground and background pixels during depth estimation

OC Stereo: Object-Centric Stereo Matching for 3D Object Detection

​

Alex D. Pon, Jason Ku, Chengyao Li, and Steven L. Waslander

International Conference on Robotics and Automation (ICRA), 2020

Object-centric stereo matching module that focuses on predicting the disparities of objects of interest to remove streaking artifacts.

BayesOD: A Bayesian Approach for Uncertainty Estimation in Deep Object Detectors

Ali Harakeh, Michael Smart, Steven L. Waslander

International Conference on Robotics and Automation (ICRA), 2020

​

An uncertainty estimation approach that reformulates the standard object detector inference and Non-Maximum suppression components from a Bayesian perspective.

VMVS: Improving 3D Object Detection for Pedestrians with Virtual Multi-View Synthesis 

​

Jason Ku, Alex D. Pon, Sean Walsh, and Steven L. Waslander

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2019

​

In this work, we presnet a virtual multi-view synthesis module for improving orientation estimation of pedestrians.

MonoPSR: Monocular 3D Object Detection Leveraging Accurate Proposals and Shape Reconstruction

Jason Ku*, Alex D. Pon*, and Steven L. Waslander

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019

Monocular 3D object detection method that uses proposals and leverages shape reconstruction. Can be found on the KITTI benchmark under MonoPSR.

AVOD:Joint 3D Proposal Generation and Object Detection from View Aggregation

Jason Ku, Melissa Mozifian, Jungwook Lee, Ali Harakeh, and Steven L. Waslander

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018

​

we propose a novel feature extractor that produces high resolution feature maps from LIDAR point clouds and RGB images, allowing for the localization of small classes in the scene.