随笔分类 -  自动驾驶

[Paper Reading] CAPE: Camera View Position Embedding for Multi-View 3D Object Detection
摘要：目录名称TL;DRMethodKey Position Embedding ConstructionQuery Position Embedding ConstructionKey/Query Position Embedding两者结合关系参考下图temporal modelingExperime 阅读全文
posted @ 2024-09-24 18:45 fariver 阅读(34) 评论(0) 推荐(0) 编辑
[Paper Reading] DEFORMABLE DETR: DEFORMABLE TRANSFORMERS FOR END-TO-END OBJECT DETECTION
摘要：DEFORMABLE DETR: DEFORMABLE TRANSFORMERS FOR END-TO-END OBJECT DETECTION link 时间：2021(ICLR) 机构：Sensetime & USTC & CUHK TL;DR 参考2D Deformable Conv，通过在R 阅读全文
posted @ 2024-08-07 17:31 fariver 阅读(36) 评论(0) 推荐(0) 编辑
[Paper Reading] DRIVEVLM: The Convergence of Autonomous Driving and Large Vision-Language Models
摘要：DRIVEVLM: The Convergence of Autonomous Driving and Large Vision-Language Models DriveVLM 时间：24.02 机构：Tsinghua University && Li Auto TL;DR 当前自动驾驶落地的主要 阅读全文
posted @ 2024-08-07 16:45 fariver 阅读(182) 评论(0) 推荐(0) 编辑
[Paper Reading] GAIA-1: A Generative World Model for Autonomous Driving
摘要：GAIA-1: A Generative World Model for Autonomous Driving GAIA-1 时间：23.09 机构：Wayve TL;DR 本文介绍一种生成世界模型，该模型利用视频、文本和动作输入来生成逼真的驾驶场景，同时提供对自身车辆行为和场景特征的细粒度控制。 阅读全文
posted @ 2024-07-03 21:19 fariver 阅读(78) 评论(0) 推荐(0) 编辑
[Paper Reading] Tesla AI Day 2022 for FSD Beta
摘要：link Framework Occupancy 模型结构比较像ICCV 2023的OccNet的做法，不过还会额外预测Suface以及NeRF state，预测可行驶区别suface的好处是可以辅助Planning&Control给出更加准确的运动速度等信息(比如，上下坡可根据suface坡度做更 阅读全文
posted @ 2024-06-11 23:49 fariver 阅读(42) 评论(0) 推荐(0) 编辑
[Paper Reading] UniAD: Planning-oriented Autonomous Driving
摘要：Planning-oriented Autonomous Driving link 时间：23.03 机构：Shanghai AI Laboratory && SenseTime TL;DR 将 感知、预测 以及 规划 模块整合成为一个E2E的网络结构。该工作是CVPR2023的Best Paper 阅读全文
posted @ 2024-06-05 20:20 fariver 阅读(402) 评论(0) 推荐(0) 编辑
[Paper Reading] MOTR: End-to-End Multiple-Object Tracking with Transformer
摘要：MOTR: End-to-End Multiple-Object Tracking with Transformer link 时间：22.07 机构：Megvii TL;DR 传统MOT通过motion与appearance来建模，有复杂的后处理难以E2E。本文基于DETR设计出MOTR算法，通过 阅读全文
posted @ 2024-05-30 23:12 fariver 阅读(348) 评论(0) 推荐(0) 编辑
[Paper Reading] FlashOcc: Fast and Memory-Efficient Occupancy Prediction via Channel-to-Height Plugin
摘要：FlashOcc: Fast and Memory-Efficient Occupancy Prediction via Channel-to-Height Plugin link 时间：23.11 机构：houmo.ai 后摩智能 TL;DR 当时比较流行的OCC方案内存与计算复杂度较高，本文提出 阅读全文
posted @ 2024-05-29 20:52 fariver 阅读(121) 评论(0) 推荐(0) 编辑
[Paper Reading] BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers
摘要：BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers link 时间：22.07 机构：Nanjing University && Sha 阅读全文
posted @ 2024-05-20 21:44 fariver 阅读(131) 评论(0) 推荐(0) 编辑
[Paper Reading] PETR: Position Embedding Transformation for Multi-View 3D Object Detection
摘要：PETR: Position Embedding Transformation for Multi-View 3D Object Detection PETR: Position Embedding Transformation for Multi-View 3D Object Detection 阅读全文
posted @ 2024-05-15 16:58 fariver 阅读(214) 评论(0) 推荐(0) 编辑
[Paper Reading] BEVDet: High-Performance Multi-Camera 3D Object Detection in Bird-Eye-View
摘要：BEVDet: High-Performance Multi-Camera 3D Object Detection in Bird-Eye-View BEVDet 时间：21/12 机构：PhiGo(鉴智机器人) TL;DR 一种BEV空间做detection的方法，构建了新颖的数据增强方法以及更新 阅读全文
posted @ 2024-05-14 14:12 fariver 阅读(125) 评论(0) 推荐(0) 编辑
[Paper Reading] OFT Orthographic Feature Transform for Monocular 3D Object Detection
摘要：OFT Orthographic Feature Transform for Monocular 3D Object Detection OFT Orthographic Feature Transform for Monocular 3D Object Detection 时间：18.11 机构： 阅读全文
posted @ 2024-05-07 21:22 fariver 阅读(107) 评论(0) 推荐(0) 编辑
[Paper Reading] LSS: Lift, Splat, Shoot: Encoding Images from Arbitrary Camera Rigs by Implicitly Unprojecting to 3D
摘要：名称 Lift, Splat, Shoot: Encoding Images from Arbitrary Camera Rigs by Implicitly Unprojecting to 3D 时间：20.08 机构：NVIDIA TL;DR 后融合方法将每一目感知结果通过相机参数转换到BEV空 阅读全文
posted @ 2024-05-06 22:58 fariver 阅读(168) 评论(0) 推荐(0) 编辑
[Paper Reading] DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queries
摘要：名称 DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queries 时间：21.10 机构：mit/CMU/Stanford TL;DR 一种利用Transformer做E2E的3D目标检测方法，在nuScenes自动 阅读全文
posted @ 2024-04-28 14:09 fariver 阅读(80) 评论(0) 推荐(0) 编辑
[基础] DETR：End-to-End Object Detection with Transformers
摘要：名称 End-to-End Object Detection with Transformers 时间：20.05 机构：Facebook AI TL;DR 文章提出一种称为DETR(Detection Transformer)的基于Transformer的检测器，相比于传统检测器不需要NMS以及a 阅读全文
posted @ 2024-04-22 22:01 fariver 阅读(90) 评论(0) 推荐(0) 编辑

点击右上角即可分享
微信分享提示