3d Object

ICCV 2023

1.3DPPE: 3D Point Positional Encoding for Transformer-based Multi-Camera 3D Object Detection

2.NeRF-Det: Learning Geometry-Aware Volumetric Representation for Multi-View 3D Object Detection以 RGB 图像为输入进行室内 3D 检测，利用 NeRF 来明确估计 3D 几何图形

3.（track）A Fast Unified System for 3D Object Detection and Tracking

4.A Simple Vision Transformer for Weakly Semi-supervised 3D Object Detection

5.（建模）Exploring Object-Centric Temporal Modeling for Efficient Multi-View 3D Object Detection 一种用于多视角三维物体检测的长序列建模框架

6.Vox-E: Text-Guided Voxel Editing of 3D Objects

7.Predict to Detect: Prediction-guided 3D Object Detection using Sequential Images 将预测方案集成到检测框架中，以明确提取和利用运动特征

8.（lidar、融合）SupFusion: Supervised LiDAR-Camera Fusion for 3D Object Detection 

9. CoIn: Contrastive Instance Feature Mining for Outdoor 3D Object Detection with Very Limited Annotations

10.（lidar、融合）ObjectFusion: Multi-modal 3D Object Detection with Object-Centric Fusion 以对象为中心将 LiDAR 点云和相机图像进行融合

11.（三维定位视觉查询）EgoLoc: Revisiting 3D Object Localization from Egocentric Videos with Visual Queries 

12.GPA-3D: Geometry-aware Prototype Alignment for Unsupervised Domain Adaptive 3D Object Detection from Point Clouds基于激光雷达的无监督域自适应三维检测框架

13.KECOR: Kernel Coding Rate Maximization for Active 3D Object Detection 主动学习（AL）

14. （自动驾驶）SA-BEV: Generating Semantic-Aware Bird's-Eye-View Feature for Multi-view 3D Object Detection语义感知 BEV 池，可以根据图像特征的语义分割过滤掉背景信息，并将图像特征转化为语义感知的 BEV 特征。

15.Delving into Motion-Aware Matching for Monocular 3D Object Tracking 单目 运动感知

16.（lidar）PG-RCNN: Semantic Surface Point Generation for 3D Object Detection 可以生成前景对象的语义表面点以进行准确检测

17.ATT3D: Amortized Text-to-3D Object Synthesis 

18.（lidar）Cross Modal Transformer: Towards Fast and Robust 3D Object Detection  将图像和点云标记作为输入，并直接输出精确的三维边界框

19.（lidar、融合）SparseFusion: Fusing Multi-Modal Sparse Representations for Multi-Sensor 3D Object Detection 利用激光雷达和相机模式中并行探测器的输出作为稀疏候选数据进行融合。

20.（lidar）DetZero: Rethinking Offboard 3D Object Detection with Long-term Sequential Point Clouds 提出了一个注意力机制提炼模块，以加强长期连续点云的上下文信息交互，从而利用分解回归方法进行物体提炼

21.Learning from Noisy Data for Semi-Supervised 3D Object Detection 设计了一个抗噪实例监督模块，以获得更好的泛化效果

22.MonoDETR: Depth-guided Transformer for Monocular 3D Object Detection 采用深度引导变换器的 DETR 单目检测框架，可以根据图像上的深度引导区域自适应地估算出查询对象三维属性

23.（建模）MonoNeRD: NeRF-like Representations for Monocular 3D Object Detection 用符号距离函数SDF对场景进行建模，采用体积渲染来恢复 RGB 图像和深度图

24.Monocular 3D Object Detection with Bounding Box Denoising in 3D by Perceiver 使用感知器 I/O 模型来融合三维到二维几何信息和二维外观信息

25.（track、lidar）TrajectoryFormer: 3D Object Tracking Transformer with Predictive Trajectory Hypotheses 

26.（lidar）PARTNER: Level up the Polar Representation for LiDAR 3D Object Detection 将点云表示为极坐标网格

27.Understanding 3D Object Interaction from a Single Image 一个基于变换器的模型，可预测物体的三维位置、物理属性和承受能力

28.Pixel-Aligned Recurrent Queries for Multi-View 3D Object Detection 编码必要的三维到二维对应关系，并捕捉输入图像的全局上下文信息

29.Zero-1-to-3: Zero-shot One Image to 3D Object 强泛化能力，从单幅图像进行三维重建

30.（lidar）Clusterformer: Cluster-based Transformer for 3D Object Detection in Point Clouds 直接借助稀疏体素特征进行查询

31.（lidar）Revisiting Domain-Adaptive 3D Object Detection by Reliable, Diverse and Class-balanced Pseudo-Labeling32.（lidar）Once Detected, Never Lost: Surpassing Human Performance in Offline LiDAR based 3D Object Detection双向跟踪模块和以轨迹为中心的学习模块

33.GraphAlign: Enhancing Accurate Feature Alignment by Graph matching for Multi-Modal 3D Object Detection

图匹配

34.（Affordance）Grounding 3D Object Affordance from 2D Interactions in Images 将不同来源的物体的区域特征统一起来，并为三维物体可承受性基础的交互情境建模

35.UpCycling: Semi-supervised 3D Object Detection without Sharing Raw-level Unlabeled Scenes 

半监督学习框架

36.（lidar）Ada3D : Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection基于体素

37.（lidar）GACE: Geometry Aware Confidence Enhancement for Black-Box 3D Object Detectors on LiDAR-Data提高检测器的置信度估计

38.QD-BEV : Quantization-aware View-guided Distillation for Multi-view 3D Object Detection 基于 BEV（鸟瞰图）的多视角三维检测，同时利用图像特征和 BEV 特征提高模型性能

39.（语言描述）Multi3DRefer: Grounding Text Description to Multiple 3D Objects 用自然语言描述在真实世界的三维场景中定位数量灵活的物体

40.ImGeoNet: Image-induced Geometry-aware Voxel Representation for Multi-view 3D Object Detection通过图像诱导的几何感知体素表示来建立三维空间模型

41.FocalFormer3D: Focusing on Hard Instance for 3D Object Detection lidarOR图像引导模型专注于挖掘困难检测实例

42.（lidar）Not Every Side Is Equal: Localization Uncertainty Estimation for Semi-Supervised 3D Object Detection半监督三维检测

43.（重建）DPF-Net: Combining Explicit Shape Priors in Deformable Primitive Field for Unsupervised Structural Reconstruction of 3D Objects

44.Towards Fair and Comprehensive Comparisons for Image-Based 3D Object Detection 建立模块化设计的代码库，制定训练方法，并讨论当前基于图像的三维物体检测方法

45.（lidar、BEV）DistillBEV: Boosting Multi-Camera 3D Object Detection with Cross-Modal Knowledge Distillation 通过模仿训练有素的基于激光雷达的教师检测器的特征来训练基于 BEV 的多摄像头学生检测器

46.（lidar、跟踪）Synchronize Feature Extracting and Matching: A Single Branch Framework for 3D Object Tracking提出了一种新颖的单分支框架 SyncTrack，将特征提取和匹配同步进行，

47.（lidar）Towards Universal LiDAR-Based 3D Object Detection by Multi-Domain Knowledge Transfer 

48.（BEV）SparseBEV: High-Performance Sparse 3D Object Detection from Multi-Camera Videos 稀疏检测器

49.（BEV）
Temporal Enhanced Training of Multi-view 3D Object Detector via Historical Object Prediction

50.（物体组合）Iterative Superquadric Recomposition of 3D Objects from Multiple Views 直接从二维视图中使用三维超四维空间作为语义部分重新组合物体，而无需训练使用三维监督的模型

51.Efficient Transformer-based 3D Object Detection with Dynamic Token Halting根据标记对检测任务的贡献，在不同层动态停止标记，从而加速基于变压器的三维物体检测器

ECCV 2022

1.DID-M3D: Decoupling Instance Depth for Monocular 3D Object Detection 单目

2.（lidar）Lidar Point Cloud Guided Monocular 3D Object Detection lidar点云引导单目检测

3.（重建）Monocular 3D Object Reconstruction with GAN Inversion 通过根据单视图观察在 3D GAN 中搜索与目标网格最相似的潜在空间来实现重建

4.（重建、增强和虚拟现实）Object Wake-Up: 3D Object Rigging from a Single Image 给定单张物体图像，通过重建其 3D 形状和骨架，以及为其合理的关节和运动设置动画来唤醒

5.（lidar与rgb）
​​​​​​Deformable Feature Aggregation for Dynamic Multi-modal 3D Object Detection 跨模态增强

6.（lidar、语义引导）Graph R-CNN: Towards Accurate 3D Object Detection with Semantic-Decorated Local Graph 融合策略来补偿语义线索有限的稀疏LiDAR点

7.Semi-Supervised Monocular 3D Object Detection by Multi-View Consistency 通过强制执行多目 3D 对象检测来改进单目 3D 对象检测视图一致性

8.Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-Training 几何对齐的多尺度训练策略

9.Monocular 3D Object Detection with Depth from Motion 利用相机自运动提供的强大几何结构来进行精确的物体深度估计和检测

10.Rethinking IoU-Based Optimization for Single-Stage 3D Object Detection 改进基于 IoU 的单阶段 3D 物体检测优化

11.DEVIANT: Depth EquiVarIAnt NeTwork for Monocular 3D Object Detection 深度等变网络

12.Label-Guided Auxiliary Training Improves 3D Object Detector 标签引导辅助增强现有 3D 对象检测器的特征学习

13.（估计深度）Densely Constrained Depth Estimator for Monocular 3D Object Detection 采用更多的投影约束并产生相当多的深度候选

14.（实时）PillarNet: Real-Time and High-Performance Pillar-Based 3D Object Detection 基于 Pillar 实时高性能检测3d

15.（lidar与rgb）Cross-Modality Knowledge Distillation Network for Monocular 3D Object Detection 跨模态知识蒸馏网络

16.（lidar）3D Object Detection with a Self-Supervised Lidar Scene Flow Backbone 自监督激光雷达场景流

17.（lidar）SWFormer: Sparse Window Transformer for 3D Object Detection in Point Clouds 

18.（lidar）FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection 全卷积网络以最少的运行时间处理大规模场景 

19.（lidar与rgb）Enhancing Multi-modal Features Using Local Self-Attention for 3D Object Detection 多模态融合

20.（track）Large-Displacement 3D Object Tracking with Hybrid Non-local Optimization 针对不同参数结合非局部和局部优化的混合方法，3d跟踪

21.（lidar）Spatially Invariant Unsupervised 3D Object-Centric Learning and Scene Decomposition 以无人监督的方式从点云中提取物体

22.PETR: Position Embedding Transformation for Multi-View 3D Object Detection 位置嵌入变换

23.（lidar与rgb融合）CramNet: Camera-Radar Fusion with Ray-Constrained Cross-Attention for Robust 3D Object Detection

24.（lidar）CenterFormer: Center-based Transformer for 3D Object Detection transformer

25.（交通场景、track）SpOT: Spatiotemporal Modeling for 3D Object Tracking 通过对对象历史的完整序列进行学习细化来改进跟踪对象的位置和运动估计

26.（lidar与rgb）Homogeneous Multi-modal Feature Fusion and Interaction for 3D Object Detection 将体素化点云特征与来自不同区域的图像特征融合

27.（lidar）Semi-Supervised 3D Object Detection with Proficient Teachers 半监督伪标签框架

28.（lidar）ProposalContrast: Unsupervised Pre-training for LiDAR-Based 3D Object Detection 无监督点云预训练框架

29.（lidar）LiDAR Distillation: Bridging the Beam-Induced Domain Gap for 3D Object Detection 

30.SpatialDETR: Robust Scalable Transformer-Based 3D Object Detection from Multi-View Camera Images with Global Cross-Sensor Attention transformer，基于每个图像内和不同视图之间的空间注意力来推断分类和边界框估计 

CVPR 2023

1.（lidar）itKD: Interchange Transfer-Based Knowledge Distillation for 3D Object Detection 

2.（VR\Affordance）Object Pop-Up: Can We Infer 3D Objects and Their Poses From Human Interactions Alone? 

3.NS3D: Neuro-Symbolic Grounding of 3D Objects and Relations 

4.GANmouflage: 3D Object Nondetection With Texture Fields 

5.MoDAR: Using Motion Forecasting for 3D Object Detection in Point Cloud Sequences 

6.OcTr: Octree-Based Transformer for 3D Object Detection

7. Semi-Supervised Stereo-Based 3D Object Detection via Cross-View Consensus 

8.