object detection资料汇总

最新推荐文章于 2025-06-25 20:01:30 发布
转载 最新推荐文章于 2025-06-25 20:01:30 发布 · 5.4k 阅读 · 14

转自：https://handong1587.github.io/deep_learning/2015/10/09/object-detection.html

Jump to...

  1. Leaderboard
  2. Papers
    1. R-CNN
    2. MultiBox
    3. SPP-Net
    4. DeepID-Net
    5. NoC
    6. Fast R-CNN
    7. DeepBox
    8. MR-CNN
    9. Faster R-CNN
    10. YOLO
    11. AttentionNet
    12. DenseBox
    13. SSD
    14. Inside-Outside Net (ION)
    15. G-CNN
    16. HyperNet
    17. MultiPathNet
    18. CRAFT
    19. OHEM
    20. R-FCN
    21. MS-CNN
    22. PVANET
    23. GBD-Net
    24. StuffNet
    25. Feature Pyramid Network (FPN)
    26. YOLOv2
    27. DSSD
    28. CC-Net
  3. Detection From Video
    1. T-CNN
    2. Datasets
  4. Object Detection in 3D
  5. Object Detection on RGB-D
  6. Salient Object Detection
    1. Saliency Detection in Video
    2. Datasets
  7. Visual Relationship Detection
  8. Specific Object Deteciton
    1. Face Deteciton
      1. UnitBox
      2. MTCNN
      3. Datasets / Benchmarks
    2. Facial Point / Landmark Detection
    3. People Detection
    4. Person Head Detection
    5. Pedestrian Detection
    6. Vehicle Detection
    7. Traffic-Sign Detection
    8. Boundary / Edge / Contour Detection
    9. Skeleton Detection
    10. Fruit Detection
    11. Part Detection
    12. Others
  9. Object Proposal
  10. Localization
  11. Tutorials / Talks
  12. Projects
  13. Blogs
MethodVOC2007VOC2010VOC2012ILSVRC 2013MSCOCO 2015Speed
OverFeat   24.3%  
R-CNN (AlexNet)58.5%53.7%53.3%31.4%  
R-CNN (VGG16)66.0%     
SPP_net(ZF-5)54.2%(1-model), 60.9%(2-model)  31.84%(1-model), 35.11%(6-model)  
DeepID-Net64.1%  50.3%  
NoC73.3% 68.8%   
Fast-RCNN (VGG16)70.0%68.8%68.4% 19.7%(@[0.5-0.95]), 35.9%(@0.5) 
MR-CNN78.2% 73.9%   
Faster-RCNN (VGG16)78.8% 75.9% 21.9%(@[0.5-0.95]), 42.7%(@0.5)198ms
Faster-RCNN (ResNet-101)85.6% 83.8% 37.4%(@[0.5-0.95]), 59.0%(@0.5) 
SSD300 (VGG16)77.2% 75.8% 25.1%(@[0.5-0.95]), 43.1%(@0.5)46 fps
SSD512 (VGG16)79.8% 78.5% 28.8%(@[0.5-0.95]), 48.5%(@0.5)19 fps
ION79.2% 76.4%   
CRAFT75.7% 71.3%48.5%  
OHEM78.9% 76.3% 25.5%(@[0.5-0.95]), 45.9%(@0.5) 
R-FCN (ResNet-50)77.4%    0.12sec(K40), 0.09sec(TitianX)
R-FCN (ResNet-101)79.5%    0.17sec(K40), 0.12sec(TitianX)
R-FCN (ResNet-101),multi sc train83.6% 82.0% 31.5%(@[0.5-0.95]), 53.2%(@0.5) 
PVANet 9.089.8% 84.2%  750ms(CPU), 46ms(TitianX)

Leaderboard

Detection Results: VOC2012

Papers

Deep Neural Networks for Object Detection

OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks

R-CNN

Rich feature hierarchies for accurate object detection and semantic segmentation

MultiBox

Scalable Object Detection using Deep Neural Networks

Scalable, High-Quality Object Detection

SPP-Net

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

DeepID-Net

DeepID-Net: Deformable Deep Convolutional Neural Networks for Object Detection

Object Detectors Emerge in Deep Scene CNNs

segDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection

NoC

Object Detection Networks on Convolutional Feature Maps

Improving Object Detection with Deep Convolutional Networks via Bayesian Optimization and Structured Prediction

Fast R-CNN

Fast R-CNN

DeepBox

DeepBox: Learning Objectness with Convolutional Networks

MR-CNN

Object detection via a multi-region & semantic segmentation-aware CNN model

Faster R-CNN

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Faster R-CNN in MXNet with distributed implementation and data parallelization

Contextual Priming and Feedback for Faster R-CNN

An Implementation of Faster RCNN with Study for Region Sampling

YOLO

You Only Look Once: Unified, Real-Time Object Detection

darkflow - translate darknet to tensorflow. Load trained weights, retrain/fine-tune them using tensorflow, export constant graph def to C++

Start Training YOLO with Our Own Data

R-CNN minus R

AttentionNet

AttentionNet: Aggregating Weak Directions for Accurate Object Detection

DenseBox

DenseBox: Unifying Landmark Localization with End to End Object Detection

SSD

SSD: Single Shot MultiBox Detector

Inside-Outside Net (ION)

Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks

Adaptive Object Detection Using Adjacency and Zoom Prediction

G-CNN

G-CNN: an Iterative Grid Based Object Detector

Factors in Finetuning Deep Model for object detection

Factors in Finetuning Deep Model for Object Detection with Long-tail Distribution

We don’t need no bounding-boxes: Training object class detectors using only human verification

HyperNet

HyperNet: Towards Accurate Region Proposal Generation and Joint Object Detection

MultiPathNet

A MultiPath Network for Object Detection

CRAFT

CRAFT Objects from Images

OHEM

Training Region-based Object Detectors with Online Hard Example Mining

Track and Transfer: Watching Videos to Simulate Strong Human Supervision for Weakly-Supervised Object Detection

Exploit All the Layers: Fast and Accurate CNN Object Detector with Scale Dependent Pooling and Cascaded Rejection Classifiers

R-FCN

R-FCN: Object Detection via Region-based Fully Convolutional Networks

Weakly supervised object detection using pseudo-strong labels

Recycle deep features for better object detection

MS-CNN

A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection

Multi-stage Object Detection with Group Recursive Learning

Subcategory-aware Convolutional Neural Networks for Object Proposals and Detection

PVANET

PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection

PVANet: Lightweight Deep Neural Networks for Real-time Object Detection

GBD-Net

Gated Bi-directional CNN for Object Detection

Crafting GBD-Net for Object Detection

StuffNet

StuffNet: Using ‘Stuff’ to Improve Object Detection

Generalized Haar Filter based Deep Networks for Real-Time Object Detection in Traffic Scene

Hierarchical Object Detection with Deep Reinforcement Learning

Learning to detect and localize many objects from few examples

Speed/accuracy trade-offs for modern convolutional object detectors

SqueezeDet: Unified, Small, Low Power Fully Convolutional Neural Networks for Real-Time Object Detection for Autonomous Driving

Feature Pyramid Network (FPN)

Feature Pyramid Networks for Object Detection

Action-Driven Object Detection with Top-Down Visual Attentions

Beyond Skip Connections: Top-Down Modulation for Object Detection

YOLOv2

YOLO9000: Better, Faster, Stronger

Yolo_mark: GUI for marking bounded boxes of objects in images for training Yolo v2

DSSD

DSSD : Deconvolutional Single Shot Detector

Wide-Residual-Inception Networks for Real-time Object Detection

Attentional Network for Visual Object Detection

CC-Net

Learning Chained Deep Features and Classifiers for Cascade in Object Detection

DeNet: Scalable Real-time Object Detection with Directed Sparse Sampling

https://arxiv.org/abs/1703.10295

A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

Spatial Memory for Context Reasoning in Object Detection

Detection From Video

Learning Object Class Detectors from Weakly Annotated Video

Analysing domain shift factors between videos and images for object detection

Video Object Recognition

Deep Learning for Saliency Prediction in Natural Video

T-CNN

T-CNN: Tubelets with Convolutional Neural Networks for Object Detection from Videos

Object Detection from Video Tubelets with Convolutional Neural Networks

Object Detection in Videos with Tubelets and Multi-context Cues

Context Matters: Refining Object Detection in Video with Recurrent Neural Networks

CNN Based Object Detection in Large Video Images

Object Detection in Videos with Tubelet Proposal Networks

Flow-Guided Feature Aggregation for Video Object Detection

Video Object Detection using Faster R-CNN

Datasets

YouTube-Objects dataset v2.2

ILSVRC2015: Object detection from video (VID)

Object Detection in 3D

Vote3Deep: Fast Object Detection in 3D Point Clouds Using Efficient Convolutional Neural Networks

Object Detection on RGB-D

Learning Rich Features from RGB-D Images for Object Detection and Segmentation

Differential Geometry Boosts Convolutional Neural Networks for Object Detection

A Self-supervised Learning System for Object Detection using Physics Simulation and Multi-view Pose Estimation

https://arxiv.org/abs/1703.03347

Salient Object Detection

This task involves predicting the salient regions of an image given by human eye fixations.

Best Deep Saliency Detection Models (CVPR 2016 & 2015)

http://i.cs.hku.hk/~yzyu/vision.html

Large-scale optimization of hierarchical features for saliency prediction in natural images

Predicting Eye Fixations using Convolutional Neural Networks

Saliency Detection by Multi-Context Deep Learning

DeepSaliency: Multi-Task Deep Neural Network Model for Salient Object Detection

SuperCNN: A Superpixelwise Convolutional Neural Network for Salient Object Detection

Shallow and Deep Convolutional Networks for Saliency Prediction

Recurrent Attentional Networks for Saliency Detection

Two-Stream Convolutional Networks for Dynamic Saliency Prediction

Unconstrained Salient Object Detection

Unconstrained Salient Object Detection via Proposal Subset Optimization

DHSNet: Deep Hierarchical Saliency Network for Salient Object Detection

Salient Object Subitizing

Deeply-Supervised Recurrent Convolutional Neural Network for Saliency Detection

Saliency Detection via Combining Region-Level and Pixel-Level Predictions with CNNs

Edge Preserving and Multi-Scale Contextual Neural Network for Salient Object Detection

A Deep Multi-Level Network for Saliency Prediction

Visual Saliency Detection Based on Multiscale Deep CNN Features

A Deep Spatial Contextual Long-term Recurrent Convolutional Network for Saliency Detection

Deeply supervised salient object detection with short connections

Weakly Supervised Top-down Salient Object Detection

SalGAN: Visual Saliency Prediction with Generative Adversarial Networks

Visual Saliency Prediction Using a Mixture of Deep Neural Networks

A Fast and Compact Salient Score Regression Network Based on Fully Convolutional Network

Saliency Detection by Forward and Backward Cues in Deep-CNNs

https://arxiv.org/abs/1703.00152

Saliency Detection in Video

Deep Learning For Video Saliency Detection

Datasets

MSRA10K Salient Object Database

http://mmcheng.net/msra10k/

Visual Relationship Detection

Visual Relationship Detection with Language Priors

ViP-CNN: A Visual Phrase Reasoning Convolutional Neural Network for Visual Relationship Detection

Visual Translation Embedding Network for Visual Relation Detection

Deep Variation-structured Reinforcement Learning for Visual Relationship and Attribute Detection

Specific Object Deteciton

Face Deteciton

Multi-view Face Detection Using Deep Convolutional Neural Networks

From Facial Parts Responses to Face Detection: A Deep Learning Approach

Compact Convolutional Neural Network Cascade for Face Detection

Face Detection with End-to-End Integration of a ConvNet and a 3D Model

CMS-RCNN: Contextual Multi-Scale Region-based CNN for Unconstrained Face Detection

Finding Tiny Faces

Towards a Deep Learning Framework for Unconstrained Face Detection

Supervised Transformer Network for Efficient Face Detection

UnitBox

UnitBox: An Advanced Object Detection Network

Bootstrapping Face Detection with Hard Negative Examples

Grid Loss: Detecting Occluded Faces

A Multi-Scale Cascade Fully Convolutional Network Face Detector

MTCNN

Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks

Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Neural Networks

Face Detection using Deep Learning: An Improved Faster RCNN Approach

Faceness-Net: Face Detection through Deep Facial Part Responses

Multi-Path Region-Based Convolutional Neural Network for Accurate Detection of Unconstrained “Hard Faces”

End-To-End Face Detection and Recognition

https://arxiv.org/abs/1703.10818

Datasets / Benchmarks

FDDB: Face Detection Data Set and Benchmark

WIDER FACE: A Face Detection Benchmark

Facial Point / Landmark Detection

Deep Convolutional Network Cascade for Facial Point Detection

Facial Landmark Detection by Deep Multi-task Learning

A Recurrent Encoder-Decoder Network for Sequential Face Alignment

Detecting facial landmarks in the video based on a hybrid framework

Deep Constrained Local Models for Facial Landmark Detection

Effective face landmark localization via single deep network

A Convolution Tree with Deconvolution Branches: Exploiting Geometric Relationships for Single Shot Keypoint Detection

https://arxiv.org/abs/1704.01880

People Detection

End-to-end people detection in crowded scenes

Detecting People in Artwork with CNNs

Deep Multi-camera People Detection

Person Head Detection

Context-aware CNNs for person head detection

Pedestrian Detection

Pedestrian Detection aided by Deep Learning Semantic Tasks

Deep Learning Strong Parts for Pedestrian Detection

Deep convolutional neural networks for pedestrian detection

Scale-aware Fast R-CNN for Pedestrian Detection

New algorithm improves speed and accuracy of pedestrian detection

Pushing the Limits of Deep CNNs for Pedestrian Detection

  • intro: “set a new record on the Caltech pedestrian dataset, lowering the log-average miss rate from 11.7% to 8.9%”
  • arxiv: http://arxiv.org/abs/1603.04525

A Real-Time Deep Learning Pedestrian Detector for Robot Navigation

A Real-Time Pedestrian Detector using Deep Learning for Human-Aware Navigation

Is Faster R-CNN Doing Well for Pedestrian Detection?

Reduced Memory Region Based Deep Convolutional Neural Network Detection

Fused DNN: A deep neural network fusion approach to fast and robust pedestrian detection

Multispectral Deep Neural Networks for Pedestrian Detection

Vehicle Detection

DAVE: A Unified Framework for Fast Vehicle Detection and Annotation

Evolving Boxes for fast Vehicle Detection

Traffic-Sign Detection

Traffic-Sign Detection and Classification in the Wild

Boundary / Edge / Contour Detection

Holistically-Nested Edge Detection

Unsupervised Learning of Edges

Pushing the Boundaries of Boundary Detection using Deep Learning

Convolutional Oriented Boundaries

Convolutional Oriented Boundaries: From Image Segmentation to High-Level Tasks

Richer Convolutional Features for Edge Detection

Skeleton Detection

Object Skeleton Extraction in Natural Images by Fusing Scale-associated Deep Side Outputs

DeepSkeleton: Learning Multi-task Scale-associated Deep Side Outputs for Object Skeleton Extraction in Natural Images

SRN: Side-output Residual Network for Object Symmetry Detection in the Wild

Fruit Detection

Deep Fruit Detection in Orchards

Image Segmentation for Fruit Detection and Yield Estimation in Apple Orchards

Part Detection

Objects as context for part detection

https://arxiv.org/abs/1703.09529

Others

Deep Deformation Network for Object Landmark Localization

Fashion Landmark Detection in the Wild

Deep Learning for Fast and Accurate Fashion Item Detection

OSMDeepOD - OSM and Deep Learning based Object Detection from Aerial Imagery (formerly known as “OSM-Crosswalk-Detection”)

Selfie Detection by Synergy-Constraint Based Convolutional Neural Network

Associative Embedding:End-to-End Learning for Joint Detection and Grouping

Deep Cuboid Detection: Beyond 2D Bounding Boxes

Automatic Model Based Dataset Generation for Fast and Accurate Crop and Weeds Detection

Deep Learning Logo Detection with Data Expansion by Synthesising Context

Pixel-wise Ear Detection with Convolutional Encoder-Decoder Networks

Automatic Handgun Detection Alarm in Videos Using Deep Learning

Object Proposal

DeepProposal: Hunting Objects by Cascading Deep Convolutional Layers

Scale-aware Pixel-wise Object Proposal Networks

Attend Refine Repeat: Active Box Proposal Generation via In-Out Localization

Learning to Segment Object Proposals via Recursive Neural Networks

Learning Detection with Diverse Proposals

  • intro: CVPR 2017
  • keywords: differentiable Determinantal Point Process (DPP) layer, Learning Detection with Diverse Proposals (LDDP)
  • arxiv: https://arxiv.org/abs/1704.03533

Localization

Beyond Bounding Boxes: Precise Localization of Objects in Images

Weakly Supervised Object Localization with Multi-fold Multiple Instance Learning

Weakly Supervised Object Localization Using Size Estimates

Active Object Localization with Deep Reinforcement Learning

Localizing objects using referring expressions

LocNet: Improving Localization Accuracy for Object Detection

Learning Deep Features for Discriminative Localization

ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization

Tutorials / Talks

Convolutional Feature Maps: Elements of efficient (and accurate) CNN-based object detection

Towards Good Practices for Recognition & Detection

Projects

TensorBox: a simple framework for training neural networks to detect objects in images

Object detection in torch: Implementation of some object detection frameworks in torch

Using DIGITS to train an Object Detection network

FCN-MultiBox Detector

KittiBox: A car detection model implemented in Tensorflow.

Blogs

Convolutional Neural Networks for Object Detection

http://rnd.azoft.com/convolutional-neural-networks-object-detection/

Introducing automatic object detection to visual search (Pinterest)

Deep Learning for Object Detection with DIGITS

Analyzing The Papers Behind Facebook’s Computer Vision Approach

Easily Create High Quality Object Detectors with Deep Learning

How to Train a Deep-Learned Object Detection Model in the Microsoft Cognitive Toolkit

Object Detection in Satellite Imagery, a Low Overhead Approach

You Only Look Twice — Multi-Scale Object Detection in Satellite Imagery With Convolutional Neural Networks

Faster R-CNN Pedestrian and Car Detection

Small U-Net for vehicle detection

【Salient Object Detection】显著性物体检测资料汇总
TomRen
03-10 1985
显著性物体检测
windows10下安装TensorFlow Object Detection API的步骤
09-19
### Windows 10 下安装 TensorFlow Object Detection API 的详细步骤 #### 概述 在 Windows 10 系统上安装并配置 TensorFlow Object Detection API 是一项技术挑战，因为该框架最初是为了 Linux 环境设计的。然而...
各种对象检测论文总结(Object Detection )
热门推荐
junmuzi的专栏
12-01 2万+
Original url: http://blog.youkuaiyun.com/u010167269/article/details/52563573 https://handong1587.github.io/deep_learning/2015/10/09/object-detection.html Object Detection  Published: 09 Oc
目标检测算法（一）——常见算法比较
疯狂挖掘机的博客
07-17 8679
一、分类：帧间差分法、背景减除法和光流法。 (1)背景减除法通过统计前若千巾贞的变化情况,从而学习背景扰动的规律。此类算法的缺点是由于通常需要缓冲若干帧频来学习背景,因此往往需要消耗大量的内存,这使其使用范围受到了限制。此外,对于大范围的背景扰动,此类算法的检测效果也不理想。Stauffer和Grimson[Stauffer99]提出的高斯混合模型是使用最为广泛的背景建模方法。高
CVPR 2024 | 基于基础模型的少样本目标检测
m0_59235945的博客
06-25 934
少样本目标检测（FSOD）旨在通过少量训练样本检测目标。视觉特征提取和查询-支持相似性学习是两个关键组件。现有的工作通常基于ImageNet预训练的视觉骨干网络，并设计复杂的度量学习网络进行少样本学习，但精度仍然较低。在本工作中，作者研究了使用现代基础模型的少样本目标检测。首先，使用仅视觉对比预训练的DINOv2模型作为视觉骨干，该模型在不调整参数的情况下表现出强大的可迁移性能。其次，使用大型语言模型（LLM）进行上下文少样本学习，输入所有类别和查询图像提议。精心设计的语言指令提示LLM在上下文中对每个提议
目标检测之序章-类别、必读论文和算法对比（实时更新）
君子之学必日新，日新者日进也。不日新者必日退，未有不进而不退者
12-21 2927
目标检测是计算机视觉领域（CV）热门且较成熟的主流研究方向之一，其应用广泛（自动驾驶、安防监控、医学影像分析、零售业货品管理等等）且已经深入到我们生活的方方面面，现在从个人学习角度，开展目标检测系列文章，主要侧重为深度学习，希望起到抛砖引玉的效果，如果有误的地方敬请指正。PS：应该注意到，近些年来，CV模型有融合的趋势，原来的目标检测模型不再是单一任务，往往能同时处理多种任务：目标检测、语义分割、实例分割、姿态检测等。
目标检测--Training Region-based Object Detectors with Online Hard Example Mining
AI小作坊 的博客
03-13 998
Training Region-based Object Detectors with Online Hard Example Miningcode： https://github.com/abhi2610/ohem本文针对 Fast R-CNN 检测系统，从样本训练的策略提出改进方法。首先来看一下 Fast R-CNN 3 Overview of Fast R-CNNFRCN（Fast R-C
在线目标检测网站
weixin_42456051的博客
05-14 4163
1、 https://ai.qq.com/product/visionimgidy.shtml#scene 登录网站后，选择多标签识别，点击本地上传，上传图片，右侧会出现结果，测试效果如下图。   2、 https://ai.baidu.com/tech/imagerecognition/object_detect 登录网站后，选择图像多主体检测，点击本地上传，上传图片，在图片右侧会出现多个主体，可以查看对不同主体的识别，测试效果如下图。 3、 https://www.studiodiip.com/
Object Detection in 20 Years A Survey总结汇报.pptx
10-27
这是Object Detection in 20 Years A Survey总结汇报，内含完整的24页制作的ppt内容，详细而又覆盖几乎所有的论文中的内容。
Google MediaPipe Objectron (3D Object Detection) 三维目标检测.zip
03-17
- `objectdetection3d_shoe_1stage.apk` 和 `objectdetection3d_shoe.apk` 专注于鞋子的检测，同样区分了一阶段和可能更精确的多阶段检测器。 - `objectdetection3d_cup.apk` 和 `objectdetection3d_camera.apk` ...
object detection 论文总结.pdf
09-13
《物体检测论文总结》 物体检测是计算机视觉领域的一个核心任务，它要求系统不仅能识别图像中的物体，还需要准确地定位其位置。R-CNN（Regions with Convolutional Neural Network Features）是这一领域的里程碑式...
目标检测实用网站
m0_60534571的博客
04-13 834
目标检测实用网站
目标检测链接
qwj的博客
03-22 373
https://handong1587.github.io/deep_learning/2015/10/09/object-detection.html#gbd-net  目标检测
目标检测GitHub网址
zhaokunhnu的博客
06-06 1366
https://www.tensorflow.org/versions/master/api_docs/python/tf/train/Saverhttps://www.tensorflow.org/versions/master/api_docs/python/tf/train/Saver
几十行代码构建一个前后端分离的目标检测演示网站，代码开源
AIZOO人工智能乐园
04-05 859
点击上方蓝色字体，关注我们在深度学习更讲究实用和落地的今天，构建一个简单的，可以利用浏览器和后端交互的演示性 Demo 可以说非常重要且实用了。本文我们将简单的介绍如何用几十行核心代码构...
目标检测web端网页代码
a15732111571的博客
05-13 1745
<!doctype html> 目标检测 1 2 3 4 5 <form enctype="multipart/form-data"> <div class="form-group"> <input i...
几个免费的在线目标检测网站
laozaoxiaowanzi的博客
05-13 8975
老师最近布置了一个作业是找一些免费的在线目标检测网站，然后室友同学找了一天没找到，我也没有，哈哈哈。本来觉得这个作业完成不了了。 然后试探性的问了下我研究nlp的高中同学，不愧是学霸，下午问的，晚上就给我找了两个。 1.https://www.studiodiip.com/2019/12/12/studio-diip-presents-computer-vision-demo/ 这个网站可以直接打开用 效果： 2.http://i-learn-machine-learning.science/
目标检测博客汇总
FengF2017的博客
08-24 489
目标检测博客汇总 文章目录目标检测博客汇总综述R-CNNSPP-NetFast R-CNNFaster R-CNNMask R-CNNYolo 对于目标检测的方法R-CNN、Fast R-CNN、Faster R-CNN、Mask R-CNN、Yolo v1-v3各选择了一篇或几篇我认为写的比较好的博客。记录一下方便以后查阅。 综述 目标检测（一）——目标检测综述 一文读懂目标检测：R-CNN...
基于Docker和Kubernetes的企业级DevOps实践训练营
Alex
07-18 1万+
基于Docker和Kubernetes的企业级DevOps实践训练营 课程准备 离线镜像包 百度：https://pan.baidu.com/s/1N1AYGCYftYGn6L0QPMWIMw 提取码：ev2h 天翼云：https://cloud.189.cn/t/ENjUbmRR7FNz CentOS7.4版本以上 虚拟机3台（4C+8G+50G），内网互通，可连外网 课件文档 《训练营课件》 《安装手册》 git仓库 https://gitee.com/agagin/python
oriented object detection与rotation object detection
最新发布
07-01
<think>我们正在讨论面向对象检测（Oriented ObjectDetection）与旋转对象检测（RotationObject Detection）技术的比较。首先需要明确这两个术语的含义，它们实际上指的是同一类技术，即检测图像中带有任意方向（旋转）的对象。在文献中，这两个术语经常互换使用，但更常见的术语是“旋转不变对象检测”或“定向对象检测”。关键点：1. **目标**：检测图像中任意方向的对象，并输出带角度的边界框（通常表示为(x,y,width,height,θ)）。2. **挑战**：传统的水平边界框检测器（如FasterR-CNN,YOLO）对旋转对象表现不佳，因为旋转会使得对象在特征图上呈现出不同的外观。主要技术方法： ###1.旋转区域提议网络（RRPN）-在FasterR-CNN框架基础上，将传统的区域提议网络（RPN）改进为旋转区域提议网络（RRPN），生成旋转的候选框（rotated proposals）。-使用角度离散化（如每15度一个bin）来预测角度偏移。-例如：Ma等人提出的RRPN（IEEETransactions onGeoscienceand RemoteSensing,2018）[^1]。###2.旋转不变特征学习-设计网络结构或损失函数，使得提取的特征对旋转具有不变性。-方法包括：-在卷积层后加入旋转不变层（如循环层、方向编码层）。-使用可变形卷积（DeformableConvolution）来适应对象的几何变换。-使用空间变换网络（STN）来对特征图进行旋转校正。 ###3.旋转敏感回归-直接回归边界框的角度参数（θ）。-为了克服角度回归的周期性难题（如0度和359度非常接近），通常采用两种方式：-将角度离散化为多个方向类别（分类问题），再结合偏移量回归。-使用角度回归的特定损失函数，如Smooth L1损失，并考虑角度的周期性（例如，将角度转换为两个分量：sin(θ)和cos(θ)的回归）。###4.基于关键点的方法-使用关键点检测框架（如CenterNet）来预测对象的中心点，然后预测宽度、高度和角度。-例如：Ding等人提出的R3Det（CVPR2021）[^2]。###5.圆形平滑标签（CSL）-为了解决角度回归的边界问题（如0度和180度相邻时的不连续问题），将角度预测视为分类问题，并使用圆形平滑标签（即相邻角度有重叠的标签分布）。-例如：Yang等人提出的CSL方法（ECCV2020）[^3]。 ###6.高斯分布表示法-将对象表示为高斯分布（均值和协方差矩阵），通过椭圆拟合来得到方向。-例如：Yang等人提出的GaussianYOLOv3（IEEE Access,2020）[^4]。 ##比较总结|方法|优点 |缺点| |------|------|------|| RRPN |直接扩展了FasterR-CNN，易于实现|计算量大，角度离散化可能导致精度损失 ||旋转不变特征 |特征对旋转鲁棒，可迁移性强|需要额外的网络结构设计，训练复杂 ||旋转敏感回归 |端到端训练，精度较高|角度回归存在边界不连续问题| |关键点方法 |简化了检测流程，速度快 |角度预测可能不够精确| |CSL|解决了角度边界问题|需要设计标签分布，计算量增加 ||高斯表示|数学表达优雅，适合任意形状 |实现复杂，后处理需要转换| ##应用场景- **遥感图像**：飞机、船只等对象具有任意方向。-**文本检测**：自然场景中的文本往往是旋转的。-**自动驾驶**：道路上的车辆等对象需要精确的方向检测。参考文献：[^1]:Ma,J.,et al. "Arbitrary-Oriented SceneText Detectionvia RotationProposals." IEEETransactions onGeoscience andRemote Sensing(2018).[^2]:Ding,J.,et al. "R3Det:RefinedSingle-Stage Detector withFeature Refinement forRotatingObject."CVPR2021. [^3]: Yang, X., etal."RethinkingtheRotation Invariance inDetection:CircularSmooth Label." ECCV2020.[^4]:Yang,X.,et al. "GaussianYOLOv3: AnAccurate andFast ObjectDetectorUsing Localization Uncertaintyfor AutonomousDriving."IEEE Access(2020).注意：上述引用文献为示例，实际文献可能需要根据具体论文调整。 §§相关问题§§1.旋转对象检测中如何处理角度回归的边界问题？2.旋转对象检测在遥感图像中有哪些具体应用？3.如何评估旋转对象检测模型的性能？常用的指标有哪些？</think>### 面向对象检测与旋转对象检测技术比较 面向对象检测（Oriented Object Detection）和旋转对象检测（Rotated Object Detection）都是针对**非水平排列对象**的检测技术，但侧重点有所不同： #### **1. 核心概念差异** - **面向对象检测** 检测具有**方向性**的对象（如车辆、飞机），输出带角度的边界框（$(x,y,w,h,\theta)$），其中$\theta$表示物体朝向（通常$-90°<\theta\leq90°$）。 **特点**：强调物体自身的朝向语义（如车头方向）[^1]。 - **旋转对象检测** 检测**任意旋转角度**的对象（如遥感图像中的建筑物），边界框表示为$(x,y,w,h,\theta)$或四边形顶点坐标。 **特点**：关注几何旋转不变性，适用于任意角度物体[^2]。 #### **2. 关键技术对比** | 技术维度 | 面向对象检测 | 旋转对象检测 | |------------------|---------------------------------------|---------------------------------------| | **边界框表示** | 旋转矩形（5参数） | 旋转矩形/四边形（5或8参数） | | **主流方法** | - R-ROI（旋转感兴趣区域）<br>- 方向敏感卷积 | - 旋转区域提议网络（RRPN）<br>- 极坐标表示 | | **损失函数** | 方向感知回归损失（如$\mathcal{L}_{angle}$） | 旋转IoU损失（解决角度周期性歧义） | | **典型应用** | 自动驾驶（车辆朝向）<br>工业质检 | 遥感图像分析<br>文档检测 | #### **3. 技术挑战** - **共享挑战** - **角度回归歧义**：$\theta$和$\theta+180°$的边界框等价性（如使用$\mathcal{L}_{smooth} = \min(|\theta-\hat{\theta}|, 180°-|\theta-\hat{\theta}|)$解决） - **特征对齐**：旋转对象导致特征图不对齐（可通过旋转RoI对齐层缓解）[^1] - **特有挑战** - 面向检测：需区分物体前后方向（如添加方向分类头） - 旋转检测：密集场景下四边形顶点易重叠（需设计顶点顺序约束） #### **4. 性能评估指标** 两类检测均使用： $$ \text{mAP}_{\theta} = \frac{1}{N}\sum_{i=1}^{N} AP_i(\theta) $$ 其中$\theta$为旋转IoU阈值（通常设0.5），但**旋转检测**额外要求： - 四边形顶点精度（如DOTA数据集指标） - 角度误差容忍度（$\Delta\theta \leq 5°$） #### **5. 典型算法** - **面向检测** - R-CNN[^1]：引入旋转RoI池化 ```python # 伪代码：旋转RoI对齐 rotated_roi = rotate(roi, angle) # 根据预测角度旋转区域 aligned_feature = bilinear_sampling(feature_map, rotated_roi) ``` - **旋转检测** - RRPN[^2]：生成旋转锚框 - CSL（Circular Smooth Label）：将角度预测转化为分类问题 --- ### 总结 | **维度** | **面向对象检测** | **旋转对象检测** | |---------------|-----------------------|-----------------------| | 核心目标 | 物体朝向语义理解 | 几何旋转不变性 | | 适用场景 | 结构化方向场景 | 任意旋转场景 | | 技术重点 | 方向分类+回归 | 旋转不变特征表示 | | 未来趋势 | 3D朝向估计融合 | 无锚点四边形预测 | ---
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值