医学图像处理中的深度学习模型

细胞病理学识别和疾病组织目标检测是目标人工智能技术在影像医学和病理方向的重要应用。 
该技术主要是前期的预处理技术复杂，主要原因是因为医学的相关病理特征成因复杂，图像方面的随机误差很大（噪音），图像断层之间的重叠。例如如果对图像的颜色没有识别要求，为了训练和计算方便采用的降维处理是将其灰化处理，然后使用分割算法将其不同的形态结构分离，该过程包含先将其腐蚀再膨胀然后过滤（一般情况中值滤波能去掉白色噪音和边缘干扰）。最后使用分类和回归来达到识别。 
在目标检测（object Detection）中常见的模型如下： 
单目标检测和多目标检测 
R-CNN 
intro: R-CNN 
arxiv: http://arxiv.org/abs/1311.2524 
supp: http://people.eecs.berkeley.edu/~rbg/papers/r-cnn-cvpr-supp.pdf 
slides: http://www.image-net.org/challenges/LSVRC/2013/slides/r-cnn-ilsvrc2013-workshop.pdf 
slides: http://www.cs.berkeley.edu/~rbg/slides/rcnn-cvpr14-slides.pdf 
github: https://github.com/rbgirshick/rcnn 
notes: http://zhangliliang.com/2014/07/23/paper-note-rcnn/ 
caffe-pr(“Make R-CNN the Caffe detection example”): https://github.com/BVLC/caffe/pull/482 
Fast R-CNN 
arxiv: http://arxiv.org/abs/1504.08083 
slides: http://tutorial.caffe.berkeleyvision.org/caffe-cvpr15-detection.pdf 
github: https://github.com/rbgirshick/fast-rcnn 
github(COCO-branch): https://github.com/rbgirshick/fast-rcnn/tree/coco 
webcam demo: https://github.com/rbgirshick/fast-rcnn/pull/29 
notes: http://zhangliliang.com/2015/05/17/paper-note-fast-rcnn/ 
notes: http://blog.csdn.net/linj_m/article/details/48930179 
github(“Fast R-CNN in MXNet”): https://github.com/precedenceguo/mx-rcnn 
github: https://github.com/mahyarnajibi/fast-rcnn-torch 
github: https://github.com/apple2373/chainer-simple-fast-rnn 
github: https://github.com/zplizzi/tensorflow-fast-rcnn 
Faster R-CNN 
intro: NIPS 2015 
arxiv: http://arxiv.org/abs/1506.01497 
gitxiv: http://www.gitxiv.com/posts/8pfpcvefDYn2gSgXk/faster-r-cnn-towards-real-time-object-detection-with-region 
slides: http://web.cs.hacettepe.edu.tr/~aykut/classes/spring2016/bil722/slides/w05-FasterR-CNN.pdf 
github(official, Matlab): https://github.com/ShaoqingRen/faster_rcnn 
github(Caffe): https://github.com/rbgirshick/py-faster-rcnn 
github(MXNet): https://github.com/msracver/Deformable-ConvNets/tree/master/faster_rcnn 
github(PyTorch–recommend): https://github.com//jwyang/faster-rcnn.pytorch 
github: https://github.com/mitmul/chainer-faster-rcnn 
github(PyTorch):: https://github.com/andreaskoepf/faster-rcnn.torch 
github(PyTorch):: https://github.com/ruotianluo/Faster-RCNN-Densecap-torch 
github(TensorFlow): https://github.com/smallcorgi/Faster-RCNN_TF 
github(TensorFlow): https://github.com/CharlesShang/TFFRCNN 
github(C++ demo): https://github.com/YihangLou/FasterRCNN-Encapsulation-Cplusplus 
github(Keras): https://github.com/yhenon/keras-frcnn 
github: https://github.com/Eniac-Xie/faster-rcnn-resnet 
github(C++): https://github.com/D-X-Y/caffe-faster-rcnn/tree/dev 
Light Head R-CNN 
A Fast R-CNN 
intro: CVPR 2017 
arxiv: https://arxiv.org/abs/1704.03414 
paper: http://abhinavsh.info/papers/pdfs/adversarial_object_detection.pdf 
github(Caffe): https://github.com/xiaolonw/adversarial-frcnn 
R-CNN minus R 
intro: BMVC 2015 
arxiv: http://arxiv.org/abs/1506.06981 
Faster R-CNN in MXNet with distributed implementation and data parallelization 
github: https://github.com/dmlc/mxnet/tree/master/example/rcnn

Contextual Priming and Feedback for Faster R-CNN 
intro: ECCV 2016. Carnegie Mellon University 
paper: http://abhinavsh.info/context_priming_feedback.pdf 
poster: http://www.eccv2016.org/files/posters/P-1A-20.pdf

An Implementation of Faster RCNN with Study for Region Sampling 
intro: Technical Report, 3 pages. CMU 
arxiv: https://arxiv.org/abs/1702.02138 
github: https://github.com/endernewton/tf-faster-rcnn

Interpretable R-CNN 
intro: North Carolina State University & Alibaba 
keywords: AND-OR Graph (AOG) 
arxiv: https://arxiv.org/abs/1711.05226

Interpretable R-CNN 
intro: North Carolina State University & Alibaba 
keywords: AND-OR Graph (AOG) 
arxiv: https://arxiv.org/abs/1711.05226 
Cascade R-CNN 
MultiBox 
SPP-Net 
Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition 
intro: ECCV 2014 / TPAMI 2015 
arxiv: http://arxiv.org/abs/1406.4729 
github: https://github.com/ShaoqingRen/SPP_net 
notes: http://zhangliliang.com/2014/09/13/paper-note-sppnet/ 
MR—CNN

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: http://www.ee.cuhk.edu.hk/%CB%9Cwlouyang/projects/imagenetDeepId/index.html 
arxiv: http://arxiv.org/abs/1412.5661

Object Detectors Emerge in Deep Scene CNNs 
intro: ICLR 2015 
arxiv: http://arxiv.org/abs/1412.6856 
paper: https://www.robots.ox.ac.uk/~vgg/rg/papers/zhou_iclr15.pdf 
paper: https://people.csail.mit.edu/khosla/papers/iclr2015_zhou.pdf 
slides: http://places.csail.mit.edu/slide_iclr2015.pdf

segDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection 
intro: CVPR 2015 
project(code+data): https://www.cs.toronto.edu/~yukun/segdeepm.html 
arxiv: https://arxiv.org/abs/1502.04275 
github: https://github.com/YknZhu/segDeepM

Object Detection Networks on Convolutional Feature Maps 
intro: TPAMI 2015 
keywords: NoC 
arxiv: http://arxiv.org/abs/1504.06066 
Improving Object Detection with Deep Convolutional Networks via 

Bayesian Optimization and Structured Prediction 
arxiv: http://arxiv.org/abs/1504.03293 
slides: http://www.ytzhang.net/files/publications/2015-cvpr-det-slides.pdf 
github: https://github.com/YutingZhang/fgs-obj

DeepBox: Learning Objectness with Convolutional Networks 
keywords: DeepBox 
arxiv: http://arxiv.org/abs/1505.02146 
github: https://github.com/weichengkuo/DeepBox

YOLO 
arxiv: http://arxiv.org/abs/1506.02640 
code: https://pjreddie.com/darknet/yolov1/ 
github: https://github.com/pjreddie/darknet 
blog: https://pjreddie.com/darknet/yolov1/ 
slides: https://docs.google.com/presentation/d/1aeRvtKG21KHdD5lg6Hgyhx5rPq_ZOsGjG5rJ1HP7BbA/pub?start=false&loop=false&delayms=3000&slide=id.p 
reddit: https://www.reddit.com/r/MachineLearning/comments/3a3m0o/realtime_object_detection_with_yolo/ 
github: https://github.com/gliese581gg/YOLO_tensorflow 
github: https://github.com/xingwangsfu/caffe-yolo 
github: https://github.com/frankzhangrui/Darknet-Yolo 
github: https://github.com/BriSkyHekun/py-darknet-yolo 
github: https://github.com/tommy-qichang/yolo.torch 
github: https://github.com/frischzenger/yolo-windows 
github: https://github.com/AlexeyAB/yolo-windows 
github: https://github.com/nilboy/tensorflow-yolo

darkflow – translate darknet to tensorflow. Load trained weights, retrain/fine-tune them using tensorflow, export constant graph def to C++ 
blog: https://thtrieu.github.io/notes/yolo-tensorflow-graph-buffer-cpp 
github: https://github.com/thtrieu/darkflow

Start Training YOLO with Our Own Data 
intro: train with customized data and class numbers/labels. Linux / Windows version for darknet. 
blog: http://guanghan.info/blog/en/my-works/train-yolo/ 
github: https://github.com/Guanghan/darknet 
YOLO: Core ML versus MPSNNGraph

intro: Tiny YOLO for iOS implemented using CoreML but also using the new MPS graph API. 
blog: http://machinethink.net/blog/yolo-coreml-versus-mps-graph/ 
github: https://github.com/hollance/YOLO-CoreML-MPSNNGraph

TensorFlow YOLO object detection on Android 
intro: Real-time object detection on Android using the YOLO network with TensorFlow 
github: https://github.com/natanielruiz/android-yolo

Computer Vision in iOS – Object Detection 
blog: https://sriraghu.com/2017/07/12/computer-vision-in-ios-object-detection/ 
github:https://github.com/r4ghu/iOS-CoreML-Yolo

YOLO2 
YOLO9000: Better, Faster, Stronger 
arxiv: https://arxiv.org/abs/1612.08242 
code: http://pjreddie.com/yolo9000/ https://pjreddie.com/darknet/yolov2/ 
github(Chainer): https://github.com/leetenki/YOLOv2 
github(Keras): https://github.com/allanzelener/YAD2K 
github(PyTorch): https://github.com/longcw/yolo2-pytorch 
github(Tensorflow): https://github.com/hizhangp/yolo_tensorflow 
github(Windows): https://github.com/AlexeyAB/darknet 
github: https://github.com/choasUp/caffe-yolo9000 
github: https://github.com/philipperemy/yolo-9000

darknet_scripts 
intro: Auxilary scripts to work with (YOLO) darknet deep learning famework. AKA -> How to generate YOLO anchors? 
github: https://github.com/Jumabek/darknet_scripts

Yolo_mark: GUI for marking bounded boxes of objects in images for training Yolo v2 
github: https://github.com/AlexeyAB/Yolo_mark

LightNet: Bringing pjreddie’s DarkNet out of the shadows 
https://github.com//explosion/lightnet

YOLO v2 Bounding Box Tool 
intro: Bounding box labeler tool to generate the training data in the format YOLO v2 requires. 
github: https://github.com/Cartucho/yolo-boundingbox-labeler-GUI

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors 
intro: LRM is the first hard example mining strategy which could fit YOLOv2 perfectly and make it better applied in series of real scenarios where both real-time rates and accurate detection are strongly demanded. 
arxiv: https://arxiv.org/abs/1804.04606 
YOLO3 
YOLOv3: An Incremental Improvement 
arxiv:https://arxiv.org/abs/1804.02767 
paper:https://pjreddie.com/media/files/papers/YOLOv3.pdf 
code: https://pjreddie.com/darknet/yolo/ 
github(Official):https://github.com/pjreddie/darknet 
github:https://github.com/experiencor/keras-yolo3 
github:https://github.com/qqwweee/keras-yolo3 
github:https://github.com/marvis/pytorch-yolo3 
github:https://github.com/ayooshkathuria/pytorch-yolo-v3 
github:https://github.com/ayooshkathuria/YOLO_v3_tutorial_from_scratch

DenseBox
SSD


• 1
• 2
• 3

• 1
• 2

intro: ECCV 2016 Oral 
arxiv: http://arxiv.org/abs/1512.02325 
paper: http://www.cs.unc.edu/~wliu/papers/ssd.pdf 
slides: http://www.cs.unc.edu/%7Ewliu/papers/ssd_eccv2016_slide.pdf 
github(Official): https://github.com/weiliu89/caffe/tree/ssd 
video: http://weibo.com/p/2304447a2326da963254c963c97fb05dd3a973 
github: https://github.com/zhreshold/mxnet-ssd 
github: https://github.com/zhreshold/mxnet-ssd.cpp 
github: https://github.com/rykov8/ssd_keras 
github: https://github.com/balancap/SSD-Tensorflow 
github: https://github.com/amdegroot/ssd.pytorch 
github(Caffe): https://github.com/chuanqi305/MobileNet-SSD 
What’s the diffience in performance between this new code you pushed and the previous code? #327 
https://github.com/weiliu89/caffe/issues/327

DSSD 
DSSD : Deconvolutional Single Shot Detector 
intro: UNC Chapel Hill & Amazon Inc 
arxiv: https://arxiv.org/abs/1701.06659 
github: https://github.com/chengyangfu/caffe/tree/dssd 
github: https://github.com/MTCloudVision/mxnet-dssd 
demo: http://120.52.72.53/www.cs.unc.edu/c3pr90ntc0td/~cyfu/dssd_lalaland.mp4

Enhancement of SSD by concatenating feature maps for object detection 
intro: rainbow SSD (R-SSD) 
arxiv: https://arxiv.org/abs/1705.09587

Context-aware Single-Shot Detector 
keywords: CSSD, DiCSSD, DeCSSD, effective receptive fields (ERFs), theoretical receptive fields (TRFs) 
arxiv: https://arxiv.org/abs/1707.08682 
Feature-Fused SSD: Fast Detection for Small Objects 
https://arxiv.org/abs/1709.05054

FSSD 
FSSD: Feature Fusion Single Shot Multibox Detector 
https://arxiv.org/abs/1712.00960

Weaving Multi-scale Context for Single Shot Detector 
intro: WeaveNet 
keywords: fuse multi-scale information 
arxiv: https://arxiv.org/abs/1712.03149 
ESSD 
Extend the shallow part of Single Shot MultiBox Detector via Convolutional Neural Network 
https://arxiv.org/abs/1801.05918

Tiny SSD: A Tiny Single-shot Detection Deep Convolutional Neural Network for Real-time Embedded Object Detection 
https://arxiv.org/abs/1802.06488

Pelee 
Pelee: A Real-Time Object Detection System on Mobile Devices 
https://github.com/Robert-JunWang/Pelee 
intro: (ICLR 2018 workshop track) 
arxiv: https://arxiv.org/abs/1804.06882 
github: https://github.com/Robert-JunWang/Pelee

R-FCN 
R-FCN: Object Detection via Region-based Fully Convolutional Networks 
arxiv: http://arxiv.org/abs/1605.06409 
github: https://github.com/daijifeng001/R-FCN 
github(MXNet): https://github.com/msracver/Deformable-ConvNets/tree/master/rfcn 
github: https://github.com/Orpine/py-R-FCN 
github: https://github.com/PureDiors/pytorch_RFCN 
github: https://github.com/bharatsingh430/py-R-FCN-multiGPU 
github: https://github.com/xdever/RFCN-tensorflow

R-FCN-3000 at 30fps: Decoupling Detection and Classification 
https://arxiv.org/abs/1712.01802

Recycle deep features for better object detection 
arxiv: http://arxiv.org/abs/1607.05066

FPN 
Feature Pyramid Networks for Object Detection 
intro: Facebook AI Research 
arxiv: https://arxiv.org/abs/1612.03144

Action-Driven Object Detection with Top-Down Visual Attentions 
arxiv: https://arxiv.org/abs/1612.06704

Beyond Skip Connections: Top-Down Modulation for Object Detection 
intro: CMU & UC Berkeley & Google Research 
arxiv: https://arxiv.org/abs/1612.06851

Wide-Residual-Inception Networks for Real-time Object Detection 
intro: Inha University 
arxiv: https://arxiv.org/abs/1702.01243

Attentional Network for Visual Object Detection 
intro: University of Maryland & Mitsubishi Electric Research Laboratories 
arxiv: https://arxiv.org/abs/1702.01478

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: https://arxiv.org/abs/1702.07054

DeNet: Scalable Real-time Object Detection with Directed Sparse Sampling 
intro: ICCV 2017 (poster) 
arxiv: https://arxiv.org/abs/1703.10295

Discriminative Bimodal Networks for Visual Localization and Detection with Natural Language Queries 
intro: CVPR 2017 
arxiv: https://arxiv.org/abs/1704.03944

Spatial Memory for Context Reasoning in Object Detection 
arxiv: https://arxiv.org/abs/1704.04224

Accurate Single Stage Detector Using Recurrent Rolling Convolution 
intro: CVPR 2017. SenseTime 
keywords: Recurrent Rolling Convolution (RRC) 
arxiv: https://arxiv.org/abs/1704.05776 
github: https://github.com/xiaohaoChen/rrc_detection

Deep Occlusion Reasoning for Multi-Camera Multi-Target Detection 
https://arxiv.org/abs/1704.05775

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: https://arxiv.org/abs/1705.05922

Point Linking Network for Object Detection 
intro: Point Linking Network (PLN) 
arxiv: https://arxiv.org/abs/1706.03646

Perceptual Generative Adversarial Networks for Small Object Detection 
https://arxiv.org/abs/1706.05274

Few-shot Object Detection 
https://arxiv.org/abs/1706.08249

Yes-Net: An effective Detector Based on Global Information 
https://arxiv.org/abs/1706.09180

SMC Faster R-CNN: Toward a scene-specialized multi-object detector 
https://arxiv.org/abs/1706.10217

Towards lightweight convolutional neural networks for object detection 
https://arxiv.org/abs/1707.01395

RON: Reverse Connection with Objectness Prior Networks for Object Detection 
intro: CVPR 2017 
arxiv: https://arxiv.org/abs/1707.01691 
github: https://github.com/taokong/RON

Mimicking Very Efficient Network for Object Detection 
intro: CVPR 2017. SenseTime & Beihang University 
paper: http://openaccess.thecvf.com/content_cvpr_2017/papers/Li_Mimicking_Very_Efficient_CVPR_2017_paper.pdf

Residual Features and Unified Prediction Network for Single Stage Detection 
https://arxiv.org/abs/1707.05031

Deformable Part-based Fully Convolutional Network for Object Detection 
intro: BMVC 2017 (oral). Sorbonne Universités & CEDRIC 
arxiv: https://arxiv.org/abs/1707.06175

Adaptive Feeding: Achieving Fast and Accurate Detections by Adaptively Combining Object Detectors 
intro: ICCV 2017 
arxiv: https://arxiv.org/abs/1707.06399

Recurrent Scale Approximation for Object Detection in CNN 
intro: ICCV 2017 
keywords: Recurrent Scale Approximation (RSA) 
arxiv: https://arxiv.org/abs/1707.09531 
github: https://github.com/sciencefans/RSA-for-object-detection 
DSOD 
DSOD: Learning Deeply Supervised Object Detectors from Scratch

img 
intro: ICCV 2017. Fudan University & Tsinghua University & Intel Labs China 
arxiv: https://arxiv.org/abs/1708.01241 
github: https://github.com/szq0214/DSOD 
github:https://github.com/Windaway/DSOD-Tensorflow 
github:https://github.com/chenyuntc/dsod.pytorch

Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids 
arxiv:https://arxiv.org/abs/1712.00886 
github:https://github.com/szq0214/GRP-DSOD

RetinaNet 
Focal Loss for Dense Object Detection 
intro: ICCV 2017 Best student paper award. Facebook AI Research 
keywords: RetinaNet 
arxiv: https://arxiv.org/abs/1708.02002

CoupleNet: Coupling Global Structure with Local Parts for Object Detection 
intro: ICCV 2017 
arxiv: https://arxiv.org/abs/1708.02863

Incremental Learning of Object Detectors without Catastrophic Forgetting 
intro: ICCV 2017. Inria 
arxiv: https://arxiv.org/abs/1708.06977

Zoom Out-and-In Network with Map Attention Decision for Region Proposal and Object Detection 
https://arxiv.org/abs/1709.04347

StairNet: Top-Down Semantic Aggregation for Accurate One Shot Detection 
https://arxiv.org/abs/1709.05788

Dynamic Zoom-in Network for Fast Object Detection in Large Images 
https://arxiv.org/abs/1711.05187

Zero-Annotation Object Detection with Web Knowledge Transfer 
intro: NTU, Singapore & Amazon 
keywords: multi-instance multi-label domain adaption learning framework 
arxiv: https://arxiv.org/abs/1711.05954

MegDet 
MegDet: A Large Mini-Batch Object Detector 
intro: Peking University & Tsinghua University & Megvii Inc 
arxiv: https://arxiv.org/abs/1711.07240

Single-Shot Refinement Neural Network for Object Detection 
arxiv: https://arxiv.org/abs/1711.06897 
github: https://github.com/sfzhang15/RefineDet

Receptive Field Block Net for Accurate and Fast Object Detection 
intro: RFBNet 
arxiv: https://arxiv.org/abs/1711.07767 
github: https://github.com//ruinmessi/RFBNet

An Analysis of Scale Invariance in Object Detection – SNIP 
arxiv: https://arxiv.org/abs/1711.08189 
github: https://github.com/bharatsingh430/snip

Feature Selective Networks for Object Detection 
https://arxiv.org/abs/1711.08879

Learning a Rotation Invariant Detector with Rotatable Bounding Box 
arxiv: https://arxiv.org/abs/1711.09405 
github: https://github.com/liulei01/DRBox

Scalable Object Detection for Stylized Objects 
intro: Microsoft AI & Research Munich 
arxiv: https://arxiv.org/abs/1711.09822

Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids 
arxiv: https://arxiv.org/abs/1712.00886 
github: https://github.com/szq0214/GRP-DSOD

Deep Regionlets for Object Detection 
keywords: region selection network, gating network 
arxiv: https://arxiv.org/abs/1712.02408

Training and Testing Object Detectors with Virtual Images 
intro: IEEE/CAA Journal of Automatica Sinica 
arxiv: https://arxiv.org/abs/1712.08470

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: https://arxiv.org/abs/1712.08832

Spot the Difference by Object Detection 
intro: Tsinghua University & JD Group 
arxiv: https://arxiv.org/abs/1801.01051

Localization-Aware Active Learning for Object Detection 
arxiv: https://arxiv.org/abs/1801.05124

Object Detection with Mask-based Feature Encoding 
https://arxiv.org/abs/1802.03934

LSTD: A Low-Shot Transfer Detector for Object Detection 
intro: AAAI 2018 
arxiv: https://arxiv.org/abs/1803.01529

Domain Adaptive Faster R-CNN for Object Detection in the Wild 
intro: CVPR 2018. ETH Zurich & ESAT/PSI 
arxiv: https://arxiv.org/abs/1803.03243

Pseudo Mask Augmented Object Detection 
https://arxiv.org/abs/1803.05858

Revisiting RCNN: On Awakening the Classification Power of Faster RCNN 
https://arxiv.org/abs/1803.06799

Zero-Shot Detection 
intro: Australian National University 
keywords: YOLO 
arxiv: https://arxiv.org/abs/1803.07113

Learning Region Features for Object Detection 
intro: Peking University & MSRA 
arxiv: https://arxiv.org/abs/1803.07066

Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection 
intro: Singapore Management University & Zhejiang University 
arxiv: https://arxiv.org/abs/1803.08208

Object Detection for Comics using Manga109 Annotations 
intro: University of Tokyo & National Institute of Informatics, Japan 
arxiv: https://arxiv.org/abs/1803.08670

Task-Driven Super Resolution: Object Detection in Low-resolution Images 
https://arxiv.org/abs/1803.11316

Transferring Common-Sense Knowledge for Object Detection 
https://arxiv.org/abs/1804.01077

Multi-scale Location-aware Kernel Representation for Object Detection 
intro: CVPR 2018 
arxiv: https://arxiv.org/abs/1804.00428 
github: https://github.com/Hwang64/MLKP

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors 
intro: National University of Defense Technology 
arxiv: https://arxiv.org/abs/1804.04606

Robust Physical Adversarial Attack on Faster R-CNN Object Detector 
https://arxiv.org/abs/1804.05810

DetNet 
DetNet: A Backbone network for Object Detection 
intro: Tsinghua University & Face++ 
arxiv: https://arxiv.org/abs/1804.06215

Other 
Relation Network for Object Detection 
intro: CVPR 2018 
arxiv: https://arxiv.org/abs/1711.11575

Quantization Mimic: Towards Very Tiny CNN for Object Detection 
Tsinghua University1 & The Chinese University of Hong Kong2 &SenseTime3 
arxiv: https://arxiv.org/abs/1805.02152 
github：https://github.com/amusi/awesome-object-detection 
Inside-OutsideNet（ION） 
CRAFI 
OHEM 
R-FCN 
MS-CNN 
PVANET 
GBD-Net 
LocNet 
等 
理论基础上数字图像的形态学处理和识别 
首先熟悉一下数字图像的定义和种类。 
数字图像：能在计算机上进行显示和操作的图像，图像是人的视觉系统对客观事物产生的视觉印象信号。按照格式常见的区分为：位图数字阵列BMP、JPG、GIF，矢量图像PNG（这种形式方便矩阵和向量（一种区别与数组的数据结构）。一般情况使用数字摄像机和数字照相机等电子设备采集的图像都认为是位图图像。 
由于时间问题今天先暂时更新这里：下面介绍一个入手项目 
仿照该项可以开发一套在线医疗检测系统

开发基础环境：tensorflow+keras来实现WEB端的多目标检测部署 
YOLO_Online 将深度学习最火的目标检测做成在线医学服务 
技术实现

web 用了 Django 来做界面，就是上传文件，保存文件这个功能。 
YOLO 的实现用的是 keras-yolo3，直接导入yolo 官方的权重即可。 
YOLO 和 web 的交互最后使用的是 socket。 
问题一、 
Django 中 Keras 初始化会有 bug，原计划是直接在 Django 里面用 keras，后来发现技术难度很大，最后用Django负责文件和socket把文件传给yolov3. 
问题二、 
最后 Django 是负责拿文件，然后用 socket 把文件名传给 yolo。 
说好的在线服务，为什么没有上线呢？买了腾讯云 1 CPU 2 G 内存，部署的时候发现 keras 根本起不来，直接被 Killed 。 
问题三、 
解决，并没有解决，因为买不起更好地服务器了，只好本地运行然后截图了。通过本地测试完成了该项目。 
django和yolo开发流程是没有把两个代码放一起，而是分开的。Django 只是上传文件，yolo 处理图片。 
项目代码：链接:https://pan.baidu.com/s/1LNv2W8EBvuHKVE6npFsaGA 密码:nj2r

YOLO 的识别是需要一定的时间的，做成 web 的服务，上传完文件之后，并不能马上识别出来，有一定的延迟。

（YOLO3），该过程希望使用合法的VPN并且开发的乌班图系统必须重英文不然文件读取各种报错和运算报错。 
搭建教程：http://www.tensorflownews.com/ 
实现步骤： 
1、从yolo官方网站下载yolo3的权重参数 
https://link.zhihu.com/?target=https%3A//pjreddie.com/media/files/yolov3.weights 
2、转换DarkNet模型位keras模型 
将python convert.py yolo3.cfg yolov3.weights model_data/yolo.h5 
3、运行yolo目标检测 
python yolo.py 
下载一张图片然后输入图的名称 
该项目的源代码地址：https://github.com/qqwweee/keras-yolo3 
该项目是Img训练的医学无法直接应用，需要通过迁移学习。 
方法就是把模型和参数迁移到你自己的数据集上训练保存一个baseline，设计一个demo