目标检测-20种常用深度学习算法、原味代码汇总

目标检测-20种常用深度学习算法、原味代码汇总

本文整理了目标检测(Object Detection)相关,20中最新的深度学习算法,以及算法相关的经典的论文和配套原味代码,分享给大家。

内容整理自:amusi/awesome-object-detection

作者:amusi


目录

· R-CNN

· Fast R-CNN

· Faster R-CNN

· Light-Head R-CNN

· Cascade R-CNN

· SPP-Net

· YOLO

· YOLOv2

· YOLOv3

· SSD

· DSSD

· FSSD

· ESSD

· Pelee

· R-FCN

· FPN

· RetinaNet

· MegDet

· DetNet

· ZSD


R-CNN

Rich feature hierarchies for accurate object detection and semantic segmentation

· intro: R-CNN

· arxiv: arxiv.org/abs/1311.2524

· supp: people.eecs.berkeley.edu

· slides: image-net.org/challenge

· slides: cs.berkeley.edu/~rbg/sl

· github: github.com/rbgirshick/r

· notes: zhangliliang.com/2014/0

· caffe-pr("Make R-CNN the Caffe detection example"): github.com/BVLC/caffe/p


Fast R-CNN

Fast R-CNN

· arxiv: arxiv.org/abs/1504.0808

· slides: tutorial.caffe.berkeleyvision.org

· github: github.com/rbgirshick/f

· github(COCO-branch): github.com/rbgirshick/f

· webcam demo: github.com/rbgirshick/f

· notes: zhangliliang.com/2015/0

· notes: blog.csdn.net/linj_m/ar

· github("Fast R-CNN in MXNet"): github.com/precedencegu

· github: github.com/mahyarnajibi

· github: github.com/apple2373/ch

· github: github.com/zplizzi/tens


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

· intro: CVPR 2017

· arxiv: arxiv.org/abs/1704.0341

· paper: abhinavsh.info/papers/p

· github(Caffe): github.com/xiaolonw/adv


Faster R-CNN

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

· intro: NIPS 2015

· arxiv: arxiv.org/abs/1506.0149

· gitxiv: gitxiv.com/posts/8pfpcv

· slides: web.cs.hacettepe.edu.tr

· github(official, Matlab): github.com/ShaoqingRen/

· github(Caffe): github.com/rbgirshick/p

· github(MXNet): github.com/msracver/Def

· github(PyTorch--recommend): github.com//jwyang/fast

· github: github.com/mitmul/chain

· github(PyTorch):: github.com/andreaskoepf

· github(PyTorch):: github.com/ruotianluo/F

· github(TensorFlow): github.com/smallcorgi/F

· github(TensorFlow): github.com/CharlesShang

· github(C++ demo): github.com/YihangLou/Fa

· github(Keras): github.com/yhenon/keras

· github: github.com/Eniac-Xie/fa

· github(C++): github.com/D-X-Y/caffe-


R-CNN minus R

· intro: BMVC 2015

· arxiv: arxiv.org/abs/1506.0698


基于MXNet,Faster R-CNN的数据并行化的分布式实现

· github: github.com/dmlc/mxnet/t

Contextual Priming and Feedback for Faster R-CNN

· intro: ECCV 2016. Carnegie Mellon University

· paper: abhinavsh.info/context_

· poster: eccv2016.org/files/post


关于Region Sampling的Faster RCNN实现

· intro: Technical Report, 3 pages. CMU

· arxiv: arxiv.org/abs/1702.0213

· github: github.com/endernewton/


可解释(Interpretable)R-CNN

· intro: North Carolina State University & Alibaba

· keywords: AND-OR Graph (AOG)

· arxiv: arxiv.org/abs/1711.0522


Light-Head R-CNN

Light-Head R-CNN: In Defense of Two-Stage Object Detector

· intro: Tsinghua University & Megvii Inc

· arxiv: arxiv.org/abs/1711.0726

· github(offical): github.com/zengarden/li

· github: github.com/terrychenism


Cascade R-CNN

Cascade R-CNN: Delving into High Quality Object Detection

· arxiv: arxiv.org/abs/1712.0072

· github: github.com/zhaoweicai/c


SPP-Net

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

· intro: ECCV 2014 / TPAMI 2015

· arxiv: arxiv.org/abs/1406.4729

· github: github.com/ShaoqingRen/

· notes: zhangliliang.com/2014/0


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

· intro: PAMI 2016

· intro: an extension of R-CNN. box pre-training, cascade on region proposals, deformation layers and context representations

· project page: ee.cuhk.edu.hk/%CB%9Cwl

· arxiv: arxiv.org/abs/1412.5661


Object Detectors Emerge in Deep Scene CNNs

· intro: ICLR 2015

· arxiv: arxiv.org/abs/1412.6856

· paper: robots.ox.ac.uk/~vgg/rg

· paper: people.csail.mit.edu/kh

· slides: places.csail.mit.edu/sl


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

· intro: CVPR 2015

· project(code+data): cs.toronto.edu/~yukun/s

· arxiv: arxiv.org/abs/1502.0427

· github: github.com/YknZhu/segDe


Object Detection Networks on Convolutional Feature Maps

· intro: TPAMI 2015

· keywords: NoC

· arxiv: arxiv.org/abs/1504.0606


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

· arxiv: arxiv.org/abs/1504.0329

· slides: ytzhang.net/files/publi

· github: github.com/YutingZhang/


DeepBox: Learning Objectness with Convolutional Networks

· keywords: DeepBox

· arxiv: arxiv.org/abs/1505.0214

· github: github.com/weichengkuo/


YOLO

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

· arxiv: arxiv.org/abs/1506.0264

· code: pjreddie.com/darknet/yo

· github: github.com/pjreddie/dar

· blog: pjreddie.com/darknet/yo

· slides: docs.google.com/present

· reddit: reddit.com/r/MachineLea

· github: github.com/gliese581gg/

· github: github.com/xingwangsfu/

· github: github.com/frankzhangru

· github: github.com/BriSkyHekun/

· github: github.com/tommy-qichan

· github: github.com/frischzenger

· github: github.com/AlexeyAB/yol

· github: github.com/nilboy/tenso

darkflow - translate darknet to tensorflow. 加载轻量级的模型,并基于Tensorflow对权重进行fine-tune,最终输出C++的constant graph。

· blog: thtrieu.github.io/notes

· github: github.com/thtrieu/dark

基于自己的数据Training YOLO

· intro: train with customized data and class numbers/labels. Linux / Windows version for darknet.

· blog: guanghan.info/blog/en/m

· github: github.com/Guanghan/dar

YOLO: Core ML versus MPSNNGraph

· intro: Tiny YOLO for iOS implemented using CoreML but also using the new MPS graph API.

· blog: machinethink.net/blog/y

· github: github.com/hollance/YOL

TensorFlow YOLO object detection on Android

· intro: Real-time object detection on Android using the YOLO network with TensorFlow

· github: github.com/natanielruiz

Computer Vision in iOS – Object Detection

· blog: sriraghu.com/2017/07/12

· github:github.com/r4ghu/iOS-Co


YOLOv2

YOLO9000: 更好,更快,更强

· arxiv: arxiv.org/abs/1612.0824

· code: pjreddie.com/yolo9000/ pjreddie.com/darknet/yo

· github(Chainer): github.com/leetenki/YOL

· github(Keras): github.com/allanzelener

· github(PyTorch): github.com/longcw/yolo2

· github(Tensorflow): github.com/hizhangp/yol

· github(Windows): github.com/AlexeyAB/dar

· github: github.com/choasUp/caff

· github: github.com/philipperemy

darknet_scripts

· intro: Auxilary scripts to work with (YOLO) darknet deep learning famework. AKA -> How to generate YOLO anchors?

· github: github.com/Jumabek/dark

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

· github: github.com/AlexeyAB/Yol

LightNet: Bringing pjreddie's DarkNet out of the shadows

github.com//explosion/l

YOLO v2 Bounding Box Tool

· intro: Bounding box labeler tool to generate the training data in the format YOLO v2 requires.

· github: github.com/Cartucho/yol

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors

· arxiv: arxiv.org/abs/1804.0460

Object detection at 200 Frames Per Second

· intro: faster than Tiny-Yolo-v2

· arXiv: arxiv.org/abs/1805.0636


YOLOv3

YOLOv3: An Incremental Improvement

· arxiv:arxiv.org/abs/1804.0276

· paper:pjreddie.com/media/file

· code: pjreddie.com/darknet/yo

· github(Official):github.com/pjreddie/dar

· github:github.com/experiencor/

· github:github.com/qqwweee/kera

· github:github.com/marvis/pytor

· github:github.com/ayooshkathur

· github:github.com/ayooshkathur


SSD

SSD: Single Shot MultiBox Detector

· intro: ECCV 2016 Oral

· arxiv: arxiv.org/abs/1512.0232

· paper: cs.unc.edu/~wliu/papers

· slides: http://www.cs.unc.edu/%7Ewliu/papers/ssd_eccv2016_slide.pdf

· github(Official): github.com/weiliu89/caf

· video: weibo.com/p/2304447a232

· github: github.com/zhreshold/mx

· github: github.com/zhreshold/mx

· github: github.com/rykov8/ssd_k

· github: github.com/balancap/SSD

· github: github.com/amdegroot/ss

· github(Caffe): github.com/chuanqi305/M

What's the diffience in performance between this new code you pushed and the previous code? #327

github.com/weiliu89/caf


DSSD

DSSD : Deconvolutional Single Shot Detector

· intro: UNC Chapel Hill & Amazon Inc

· arxiv: arxiv.org/abs/1701.0665

· github: github.com/chengyangfu/

· github: github.com/MTCloudVisio

· demo: 120.52.72.53/http://www

Enhancement of SSD by concatenating feature maps for object detection

· intro: rainbow SSD (R-SSD)

· arxiv: arxiv.org/abs/1705.0958

Context-aware Single-Shot Detector

· keywords: CSSD, DiCSSD, DeCSSD, effective receptive fields (ERFs), theoretical receptive fields (TRFs)

· arxiv: arxiv.org/abs/1707.0868

Feature-Fused SSD: Fast Detection for Small Objects

arxiv.org/abs/1709.0505


FSSD

FSSD: Feature Fusion Single Shot Multibox Detector

arxiv.org/abs/1712.0096

Weaving Multi-scale Context for Single Shot Detector

· intro: WeaveNet

· keywords: fuse multi-scale information

· arxiv: arxiv.org/abs/1712.0314


ESSD

Extend the shallow part of Single Shot MultiBox Detector via Convolutional Neural Network

arxiv.org/abs/1801.0591

Tiny SSD: A Tiny Single-shot Detection Deep Convolutional Neural Network for Real-time Embedded Object Detection

arxiv.org/abs/1802.0648


Pelee

Pelee: A Real-Time Object Detection System on Mobile Devices

github.com/Robert-JunWa

intro: (ICLR 2018 workshop track)

arxiv: arxiv.org/abs/1804.0688

github: github.com/Robert-JunWa


R-FCN

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

· arxiv: arxiv.org/abs/1605.0640

· github: github.com/daijifeng001

· github(MXNet): github.com/msracver/Def

· github: github.com/Orpine/py-R-

· github: github.com/PureDiors/py

· github: github.com/bharatsingh4

· github: github.com/xdever/RFCN-

R-FCN-3000 at 30fps: Decoupling Detection and Classification

arxiv.org/abs/1712.0180

Recycle deep features for better object detection

· arxiv: arxiv.org/abs/1607.0506


FPN

Feature Pyramid Networks for Object Detection

· intro: Facebook AI Research

· arxiv: arxiv.org/abs/1612.0314

Action-Driven Object Detection with Top-Down Visual Attentions

· arxiv: arxiv.org/abs/1612.0670

Beyond Skip Connections: Top-Down Modulation for Object Detection

· intro: CMU & UC Berkeley & Google Research

· arxiv: arxiv.org/abs/1612.0685

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

· intro: Inha University

· arxiv: arxiv.org/abs/1702.0124

Attentional Network for Visual Object Detection

· intro: University of Maryland & Mitsubishi Electric Research Laboratories

· arxiv: arxiv.org/abs/1702.0147

Learning Chained Deep Features and Classifiers for Cascade in Object Detection

· keykwords: CC-Net

· intro: chained cascade network (CC-Net). 81.1% mAP on PASCAL VOC 2007

· arxiv: arxiv.org/abs/1702.0705

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

· intro: ICCV 2017 (poster)

· arxiv: arxiv.org/abs/1703.1029

Discriminative Bimodal Networks for Visual Localization and Detection with Natural Language Queries

· intro: CVPR 2017

· arxiv: arxiv.org/abs/1704.0394

Spatial Memory for Context Reasoning in Object Detection

· arxiv: arxiv.org/abs/1704.0422

Accurate Single Stage Detector Using Recurrent Rolling Convolution

· intro: CVPR 2017. SenseTime

· keywords: Recurrent Rolling Convolution (RRC)

· arxiv: arxiv.org/abs/1704.0577

· github: github.com/xiaohaoChen/

Deep Occlusion Reasoning for Multi-Camera Multi-Target Detection

arxiv.org/abs/1704.0577

LCDet: Low-Complexity Fully-Convolutional Neural Networks for Object Detection in Embedded Systems

· intro: Embedded Vision Workshop in CVPR. UC San Diego & Qualcomm Inc

· arxiv: arxiv.org/abs/1705.0592

Point Linking Network for Object Detection

· intro: Point Linking Network (PLN)

· arxiv: arxiv.org/abs/1706.0364

Perceptual Generative Adversarial Networks for Small Object Detection

arxiv.org/abs/1706.0527

Few-shot Object Detection

arxiv.org/abs/1706.0824

Yes-Net: An effective Detector Based on Global Information

arxiv.org/abs/1706.0918

SMC Faster R-CNN: Toward a scene-specialized multi-object detector

arxiv.org/abs/1706.1021

Towards lightweight convolutional neural networks for object detection

arxiv.org/abs/1707.0139

RON: Reverse Connection with Objectness Prior Networks for Object Detection

· intro: CVPR 2017

· arxiv: arxiv.org/abs/1707.0169

· github: github.com/taokong/RON

Mimicking Very Efficient Network for Object Detection

· intro: CVPR 2017. SenseTime & Beihang University

· paper: openaccess.thecvf.com/c

Residual Features and Unified Prediction Network for Single Stage Detection

arxiv.org/abs/1707.0503

Deformable Part-based Fully Convolutional Network for Object Detection

· intro: BMVC 2017 (oral). Sorbonne Universités & CEDRIC

· arxiv: arxiv.org/abs/1707.0617

Adaptive Feeding: Achieving Fast and Accurate Detections by Adaptively Combining Object Detectors

· intro: ICCV 2017

· arxiv: arxiv.org/abs/1707.0639

Recurrent Scale Approximation for Object Detection in CNN

· intro: ICCV 2017

· keywords: Recurrent Scale Approximation (RSA)

· arxiv: arxiv.org/abs/1707.0953

· github: github.com/sciencefans/


DSOD

DSOD: Learning Deeply Supervised Object Detectors from Scratch

· intro: ICCV 2017. Fudan University & Tsinghua University & Intel Labs China

· arxiv: arxiv.org/abs/1708.0124

· github: github.com/szq0214/DSOD

· github:github.com/Windaway/DSO

· github:github.com/chenyuntc/ds

Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids

· arxiv:arxiv.org/abs/1712.0088

· github:github.com/szq0214/GRP-


RetinaNet

Focal Loss for Dense Object Detection

· intro: ICCV 2017 Best student paper award. Facebook AI Research

· keywords: RetinaNet

· arxiv: arxiv.org/abs/1708.0200

CoupleNet: Coupling Global Structure with Local Parts for Object Detection

· intro: ICCV 2017

· arxiv: arxiv.org/abs/1708.0286

Incremental Learning of Object Detectors without Catastrophic Forgetting

· intro: ICCV 2017. Inria

· arxiv: arxiv.org/abs/1708.0697

Zoom Out-and-In Network with Map Attention Decision for Region Proposal and Object Detection

arxiv.org/abs/1709.0434

StairNet: Top-Down Semantic Aggregation for Accurate One Shot Detection

arxiv.org/abs/1709.0578

Dynamic Zoom-in Network for Fast Object Detection in Large Images

arxiv.org/abs/1711.0518

Zero-Annotation Object Detection with Web Knowledge Transfer

· intro: NTU, Singapore & Amazon

· keywords: multi-instance multi-label domain adaption learning framework

· arxiv: arxiv.org/abs/1711.0595


MegDet

MegDet: A Large Mini-Batch Object Detector

· arxiv: arxiv.org/abs/1711.0724


Single-Shot Refinement Neural Network for Object Detection

· arxiv: arxiv.org/abs/1711.0689

· github: github.com/sfzhang15/Re


Receptive Field Block Net for Accurate and Fast Object Detection

· arxiv: arxiv.org/abs/1711.0776

· github: github.com//ruinmessi/R


An Analysis of Scale Invariance in Object Detection - SNIP

· arxiv: arxiv.org/abs/1711.0818

· github: github.com/bharatsingh4


Feature Selective Networks for Object Detection

arxiv.org/abs/1711.0887


Learning a Rotation Invariant Detector with Rotatable Bounding Box

· arxiv: arxiv.org/abs/1711.0940

· github: github.com/liulei01/DRB


Scalable Object Detection for Stylized Objects

· intro: Microsoft AI & Research Munich

· arxiv: arxiv.org/abs/1711.0982


Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids

· arxiv: arxiv.org/abs/1712.0088

· github: github.com/szq0214/GRP-


Deep Regionlets for Object Detection

· keywords: region selection network, gating network

· arxiv: arxiv.org/abs/1712.0240

Training and Testing Object Detectors with Virtual Images

· intro: IEEE/CAA Journal of Automatica Sinica

· arxiv: arxiv.org/abs/1712.0847

Large-Scale Object Discovery and Detector Adaptation from Unlabeled Video

· keywords: object mining, object tracking, unsupervised object discovery by appearance-based clustering, self-supervised detector adaptation

· arxiv: arxiv.org/abs/1712.0883

Spot the Difference by Object Detection

· intro: Tsinghua University & JD Group

· arxiv: arxiv.org/abs/1801.0105

Localization-Aware Active Learning for Object Detection

· arxiv: arxiv.org/abs/1801.0512

Object Detection with Mask-based Feature Encoding

arxiv.org/abs/1802.0393

LSTD: A Low-Shot Transfer Detector for Object Detection

· intro: AAAI 2018

· arxiv: arxiv.org/abs/1803.0152

Domain Adaptive Faster R-CNN for Object Detection in the Wild

· intro: CVPR 2018. ETH Zurich & ESAT/PSI

· arxiv: arxiv.org/abs/1803.0324

Pseudo Mask Augmented Object Detection

arxiv.org/abs/1803.0585

Revisiting RCNN: On Awakening the Classification Power of Faster RCNN

arxiv.org/abs/1803.0679

Zero-Shot Detection

· intro: Australian National University

· keywords: YOLO

· arxiv: arxiv.org/abs/1803.0711

Learning Region Features for Object Detection

· intro: Peking University & MSRA

· arxiv: arxiv.org/abs/1803.0706

Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection

· intro: Singapore Management University & Zhejiang University

· arxiv: arxiv.org/abs/1803.0820

Object Detection for Comics using Manga109 Annotations

· intro: University of Tokyo & National Institute of Informatics, Japan

· arxiv: arxiv.org/abs/1803.0867

Task-Driven Super Resolution: Object Detection in Low-resolution Images

arxiv.org/abs/1803.1131

Transferring Common-Sense Knowledge for Object Detection

arxiv.org/abs/1804.0107

Multi-scale Location-aware Kernel Representation for Object Detection

· intro: CVPR 2018

· arxiv: arxiv.org/abs/1804.0042

· github: github.com/Hwang64/MLKP

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors

· intro: National University of Defense Technology

· arxiv: arxiv.org/abs/1804.0460

Robust Physical Adversarial Attack on Faster R-CNN Object Detector

arxiv.org/abs/1804.0581


DetNet

DetNet: A Backbone network for Object Detection

arxiv: arxiv.org/abs/1804.0621


LMNet: Real-time Multiclass Object Detection on CPU using 3D LiDARs

· arxiv: arxiv.org/abs/1805.0490

· github: github.com/CPFL/Autowar


ZSD

Zero-Shot Object Detection

· arxiv: arxiv.org/abs/1804.0434


Zero-Shot Object Detection: Learning to Simultaneously Recognize and Localize Novel Concepts

· arxiv: arxiv.org/abs/1803.0604


Zero-Shot Object Detection by Hybrid Region Embedding

· arxiv: arxiv.org/abs/1805.0615

发布于 2018-05-19

文章被以下专栏收录

    专注深度学习、NLP相关技术、资讯,追求纯粹的技术,享受学习、分享的快乐。欢迎扫描头像二维码或者微信搜索“深度学习与NLP”公众号添加关注,获得更多深度学习与NLP方面的经典论文、实践经验和最新消息。