3D Object Detection

515 papers with code • 54 benchmarks • 48 datasets

3D Object Detection is a task in computer vision where the goal is to identify and locate objects in a 3D environment based on their shape, location, and orientation. It involves detecting the presence of objects and determining their location in the 3D space in real-time. This task is crucial for applications such as autonomous vehicles, robotics, and augmented reality.

( Image credit: AVOD )

Benchmarks

Add a Result

These leaderboards are used to track progress in 3D Object Detection

Dataset	Best Model	Compare
KITTI Cars Moderate	GLENet-VR	See all
nuScenes	EA-LSS	See all
SUN-RGBD val	Point-GCC+TR3D+FF	See all
ScanNetV2	V-DETR	See all
KITTI Cars Easy	GLENet-VR	See all
KITTI Cars Hard	3D Dual-Fusion	See all
nuScenes Camera Only	Far3D	See all
KITTI Pedestrians Moderate	3D-FCT	See all
KITTI Cyclists Easy	3D-FCT	See all
KITTI Cyclists Moderate	3D-FCT	See all
KITTI Cyclists Hard	SA-Det3D	See all
KITTI Cars Easy val	SA-SSD+EBM	See all
KITTI Cars Moderate val	SA-SSD+EBM	See all
KITTI Cars Hard val	M3DeTR	See all
KITTI Pedestrians Easy	IPOD	See all
KITTI Pedestrians Hard	SVGA-Net	See all
waymo vehicle	PillarNeXt	See all
DAIR-V2X-I	CoBEV	See all
SUN-RGBD	CAGroup3D (Geo Only)	See all
waymo cyclist	DSVT(val)	See all
waymo pedestrian	DSVT(val)	See all
Rope3D	CoBEV	See all
V2XSet	V2X-ViT	See all
nuScenes LiDAR only	DSVT	See all
S3DIS	Point-GCC+TR3D	See all
OPV2V	V2VNet (PointPillar backbone)	See all
KITTI Pedestrian Easy val	PVCNN	See all
KITTI Pedestrian Moderate val	PVCNN	See all
KITTI Pedestrian Hard val	PVCNN	See all
KITTI Cyclist Easy val	M3DeTR	See all
KITTI Cyclist Moderate val	M3DeTR	See all
KITTI Cyclist Hard val	M3DeTR	See all
aiMotive Dataset	Lidar-Radar-Camera	See all
3D Object Detection on Argoverse2 Camera Only	Far3D	See all
waymo all_ns	CenterPoint	See all
DAIR-V2X	CoBEVFlow	See all
NYU Depth v2	SGPN-CNN	See all
nuScenes-F	RRPN + R101 - F	See all
nuScenes-FB	RRPN + R101 - FB	See all
KITTI Pedestrian Hard	PiFeNet	See all
KITTI Cyclists Moderate val	Deformable PV-RCNN	See all
KITTI Pedestrians Moderate val	Deformable PV-RCNN	See all
Dense Fog	PV-RCNN	See all
KITTI Pedestrian Moderate	PiFeNet	See all
Heavy Snowfall	PV-RCNN	See all
Light Snowfall	PV-RCNN	See all
Clear Weather	PV-RCNN	See all
KITTI Pedestrian Easy	PiFeNet	See all
KITTI Pedestrian	PiFeNet	See all
V2X-SIM	Where2comm	See all
Argoverse2	VoxelNeXt	See all
Cityscapes 3D	TaskPrompter	See all
IRV2V	CoBEVFlow	See all
Argoverse	ky_nctu_mo	See all

Show all 54 benchmarks

Collapse benchmarks

Libraries

Use these libraries to find 3D Object Detection models and implementations

open-mmlab/mmdetection3d

13 papers

4,311

PaddlePaddle/Paddle3D

6 papers

479

open-mmlab/OpenPCDet

5 papers

4,012

DerrickXuNu/OpenCOOD

5 papers

545

See all 7 libraries.

Datasets

Subtasks

Robust 3D Object Detection

Robust BEV Detection

Most implemented papers

Most implemented Social Latest No code

YOLO9000: Better, Faster, Stronger

AlexeyAB/darknet • • CVPR 2017

On the 156 classes not in COCO, YOLO9000 gets 16. 0 mAP.

234

Paper
Code

Frustum PointNets for 3D Object Detection from RGB-D Data

charlesq34/frustum-pointnets • • CVPR 2018

In this work, we study 3D object detection from RGB-D data in both indoor and outdoor scenes.

Paper
Code

VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection

qianguih/voxelnet • • CVPR 2018

Accurate detection of objects in 3D point clouds is a central problem in many applications, such as autonomous navigation, housekeeping robots, and augmented/virtual reality.

Paper
Code

nuScenes: A multimodal dataset for autonomous driving

nutonomy/nuscenes-devkit • CVPR 2020

Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar.

Paper
Code

PointPillars: Fast Encoders for Object Detection from Point Clouds

nutonomy/second.pytorch • • CVPR 2019

These benchmarks suggest that PointPillars is an appropriate encoding for object detection in point clouds.

Paper
Code

Deep Hough Voting for 3D Object Detection in Point Clouds

facebookresearch/votenet • • ICCV 2019

Current 3D object detection methods are heavily influenced by 2D detectors.

Paper
Code

3D Bounding Box Estimation Using Deep Learning and Geometry

smallcorgi/3D-Deepbox • • CVPR 2017

In contrast to current techniques that only regress the 3D orientation of an object, our method first regresses relatively stable 3D object properties using a deep convolutional neural network and then combines these estimates with geometric constraints provided by a 2D object bounding box to produce a complete 3D bounding box.

Paper
Code