Introduction and download of common data sets for in-depth learning (with network disk link)

 

Abstract

This blog post summarizes the data sets commonly used by bloggers for in-depth learning , Contains commonly used classifications 、 Target detection and face recognition tasks , A brief introduction to each data set is given 、 Keywords obtained from the official website download website and official account . Because some data sets are large , The download speed of the official website may be slow , For the convenience of management , Here I put it into the personal official account platform , Available through search “AI Technology research and sharing ” Official account , You can get the network disk link by replying to the keywords of each data set in the background . Data set files are downloaded from the official website , Only for learning communication , Bloggers will continue to update , Welcome to your attention .

---

Catalog

Abstract

1. Classification data set

Data sets ：MNIST

Data sets ：CIFAR-10

Data sets ：Fashion-MNIST

Data sets ：Imagenet

2. Target detection data set

Data sets ：MS-COCO

Data sets ：PASCAL-VOC

Data sets ：BDD100K

Data sets ：Open-Images

3. Face recognition dataset

Data sets ：CASIA-WebFace

Data sets ：PubFig

Data sets ：CelebA

Data sets ：ColorFeret

Data sets ：MTFL

Data sets ：FaceDB

Data sets ：LFW

Data sets ：Person identification in TV series

Data sets ：CMUVASC-PIE

Data sets ：CASIA-FaceV5

Data sets ：The CNBC Face Database

Data sets ：IMDB-WIKI

Data sets ：FDDB

Data sets ：Caltech-10K-WebFaces

Data sets ：JAFFE

Data sets ：AFLW

---

1. Classification data set

	Data sets ：MNIST Handwritten digital datasets , Contains a set of 60,000 A training set of examples and a set of 10,000 A test set of samples .
Official website address	http://yann.lecun.com/exdb/mnist/
keyword （ It is recommended to copy ）	Mnist

	Data sets ：CIFAR-10 Contains 10 Category 60,000 Images （ Each class is represented as a line in the figure above ）. All in all 50,000 A training image and 10,000 A test image .
Official website address	http://www.cs.toronto.edu/~kriz/cifar.html
keyword （ It is recommended to copy ）	CIFAR-10

	Data sets ：Fashion-MNIST contain 60,000 A training image and 10,000 A test image . similar MNIST Fashion products database of .
Official website address	https://github.com/zalandoresearch/fashion-mnist
keyword （ It is recommended to copy ）	fashion-mnist

	Data sets ：Imagenet Imagenet The dataset has 1400 More than ten thousand pictures , cover 2 More than 10000 categories , About image classification 、 location 、 Detection and other research work are mostly based on this data set .
Official website address	http://www.image-net.org/about-stats
keyword （ It is recommended to copy ）	ImageNet

 

2. Target detection data set

	Data sets ：MS-COCO COCO It is a large-scale and rich object detection , Segmentation and subtitle data sets .330K Images ,80 Object categories , Each image 5 Subtitles ,250,000 A person with key points .
Official website address	http://cocodataset.org/
keyword （ It is recommended to copy ）	MS-COCO

	Data sets ：PASCAL-VOC PASCAL VOC Challenge is a benchmark of visual object classification, recognition and detection , It provides standard image annotation data set and standard evaluation system for detection algorithm and learning performance , Include 20 A catalog .
Official website address	http://host.robots.ox.ac.uk/pascal/VOC/
keyword （ It is recommended to copy ）	PASCAL-VOC

	Data sets ：BDD100K Autonomous driving is commonly used in large and diverse data sets , Dimension exceeds 100,000 Zhang image , The category contains buses , Pedestrians , Bicycle , truck , The car , Trains and riders , For target detection 、 Full frame segmentation, etc .
Official website address	http://bdd-data.berkeley.edu/
keyword （ It is recommended to copy ）	BDD-100K

	Data sets ：Open-Images Open Images Is a containing near 900 Ten thousand images URL Data set of . These images have been annotated with image level label borders for thousands of categories .
Official website address	https://storage.googleapis.com/openimages/web/index.html
keyword （ It is recommended to copy ）	Open-Images

 

3. Face recognition dataset

	Data sets ：CASIA-WebFace Contains 10575 Individual 494414 Zhang image .CASIA-webface database , It contains 10000 personal , altogether 50 Ten thousand face pictures , All from the network .
Official website address	http://en.findeen.com/casia_webface.html
keyword （ It is recommended to copy ）	CASIA-WebFace

	Data sets ：PubFig Columbia University's public face data set , contains 200 Individual 58k+ Face image , It is mainly used for face recognition in unrestricted scenes .
Official website address	http://www.cs.columbia.edu/CAVE/databases/pubfig/
keyword （ It is recommended to copy ）	PubFig

	Data sets ：CelebA The large face recognition data set published by Professor Tang Xiaoou's Laboratory of the Chinese University of Hong Kong . The dataset contains 200K Face pictures , Face attributes include 40 Varied , It is mainly used for face attribute recognition .
Official website address	http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
keyword （ It is recommended to copy ）	CelebA

	Data sets ：ColorFeret It includes a general face database and a general test standard , It already contains 1000 Many people 10000 Multiple photos , Everyone includes different expressions , light , Photos of posture and age .
Official website address	https://www.nist.gov/itl/iad/image-group/color-feret-database
keyword （ It is recommended to copy ）	ColorFeret

	Data sets ：MTFL The dataset contains nearly 13000 Face pictures , All from the network .
Official website address	http://mmlab.ie.cuhk.edu.hk/projects/TCDCN.html
keyword （ It is recommended to copy ）	MTFL

	Data sets ：FaceDB Contains 1521 The amplitude resolution is 384x286 Pixel gray image . Each image comes from 23 A different tester's frontal angle of the face .
Official website address	https://www.bioid.com/About/BioID-Face-Database
keyword （ It is recommended to copy ）	FaceDB

	Data sets ：LFW In order to study the problem of face recognition in unrestricted environment . This dataset contains more than 13,000 Face images , All collected in Internet. In order to study the problem of face recognition in unrestricted environment . This dataset contains more than 13,000 Face images , All collected in Internet.
Official website address	http://vis-www.cs.umass.edu/lfw/
keyword （ It is recommended to copy ）	LFW

	Data sets ：Person identification in TV series The face photos selected in this dataset are from two well-known TV dramas ,《 Vampire Hunter Buffy 》 and 《 The big bang theory 》.
Official website address	https://cvhci.anthropomatik.kit.edu/~baeuml/publications/semi-supervised-learning-with-constraints-for-person-identification-in-multimedia-data/
keyword （ It is recommended to copy ）	PITVS

	Data sets ：CMUVASC-PIE CMU PIE The face database is built on 2000 year 11 month , It includes information from 68 Individual 40000 A picture , It includes everyone's 13 Kind of posture conditions ,43 Light conditions and 4 A picture with an expression .
Official website address	http://vasc.ri.cmu.edu/idb/html/face/index.html
keyword （ It is recommended to copy ）	CMUVASC-PIE

	Data sets ：CASIA-FaceV5 The dataset contains data from 500 Individual 2500 Zhang Yaya's face .
Official website address	http://biometrics.idealtest.org/dbDetailForUser.do?id=9
keyword （ It is recommended to copy ）	CASIA-FaceV5

	Data sets ：The CNBC Face Database This data set collects 200 Individuals in different states （ Different looks , Dress up , Hair style, etc ） Face photos of .
Official website address	http://wiki.cnbc.cmu.edu/Face_Place
keyword （ It is recommended to copy ）	Face-Place

	Data sets ：IMDB-WIKI IMDB-WIKI The face database totals 523,051 Zhang face database , Each picture is marked with the person's age and gender , It is of great significance to the study of age recognition and gender recognition .
Official website address	https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/
keyword （ It is recommended to copy ）	IMDB-WIKI

	Data sets ：FDDB FDDB yes UMass Data set of , Used for human face detection (Face Detection). This data set is relatively large , It's challenging .
Official website address	http://vis-www.cs.umass.edu/fddb/index.html
keyword （ It is recommended to copy ）	FDDB

	Data sets ：Caltech-10K-WebFaces 10k+ Face , Provide coordinate positions of eyes and mouth
Official website address	http://www.vision.caltech.edu/Image_Datasets/Caltech_10K_WebFaces/#Description
keyword （ It is recommended to copy ）	Caltech-10K

	Data sets ：JAFFE The database is created by 10 Japanese women made various expressions according to instructions in the experimental environment . The whole database has 213 Zhang image ,10 personal , It's all women , Everyone makes 7 Kind of expression .
Official website address	http://www.kasrl.org/jaffe.html
keyword （ It is recommended to copy ）	JAFFE

	Data sets ：AFLW front 2000 individual AFLW Fitting of samples 3D Noodles , Can be used for 3D Face the assessment .
Official website address	http://www.cbsr.ia.ac.cn/users/xiangyuzhu/projects/3DDFA/main.htm
keyword （ It is recommended to copy ）	AFLW

By searching “AI Technology research and sharing ” Official account , And in The background replies to each dataset keyword You can get the network disk link . Data set files are downloaded from the official website , Only for learning communication , Bloggers will continue to update , Welcome to your attention .

Share more , Coming soon ！