当前位置:网站首页>More than 50 object detection datasets from different industries

More than 50 object detection datasets from different industries

2020-11-07 20:55:42 Artificial intelligence meets pioneer

author |Abhishek Annamraju
compile |Flin
source |medium

Computer vision is a rapidly developing field , A large number of new technologies and algorithms appear in different conferences and journals every day . Speaking of target detection , In theory, you'll learn a lot of algorithms , such as Faster-rcnn、Mask rcnn、Yolo、SSD、Retinenet、 cascade rcnn、Peleenet、EfficientDet、CornerNet…. This algorithm list is never finished !

It is always beneficial to consolidate your learning experience by applying it to different data sets !!!

thus , You tend to understand algorithms better , And we can intuitively understand which algorithms can run on which dataset .

We are Monk Computer Vision Org The open source team at the , List of image segmentation and action recognition data sets , A short tutorial is created for each object , You can use these datasets and try different object detection algorithms

The following is a short list of object detection datasets , Brief details about them and how to use them . Data sets come from the following areas :

* Agriculture
* Advanced driver assistance and automatic driving vehicle system
* fashion , Retail and marketing
* Wild animal
* sports
* Satellite imaging
* Medical imaging
* Security and surveillance
* Underwater imaging

….. And more !!!!!

The complete list can be found in github Find the relevant instructions and training code on

Data sets related to agriculture

A)Winegrape Test data set

* The goal is : Detection of grape clusters in vineyards

* application : Monitor growth and analyze yield

* Details :300 Images , with 5 Grape varieties 4400 A bounding box

* How to utilize data sets and use YoloV3 Pipeline build custom detector

B) Global wheat detection data set

* The goal is : Testing wheat crops in the field

* application : Monitor growth and analyze yield

* Details : with 100K + endorsed 3430 Images

* How to utilize data sets and use EfficientDet-D4 Pipeline build custom detector

Advanced data set for driver assistance and autopilot vehicle systems

A)LISA Traffic sign detection data set

* The goal is : It is used to detect and classify traffic signs in dash cam images

* application : Traffic sign recognition is a rule setting program for automatic driving

* Details : stay 47 On the type of American logo 6610 There are on the frame 7855 A note

* How to make use of data set and establish custom usage EfficientDet-D3 Detector of pipeline

* Another dataset is added to this repository

B) Object detection under low illumination

* The goal is : Detection of objects on the road under low light conditions —— Fog , smog , Rain, etc

* application : This is an important part of autopilot. , Because it can detect objects , Therefore, it is a safer vehicle under adverse conditions

* details : stay 12 On different object types 7500 On frame 15K + notes

* How to utilize data sets and use EfficientDet-D3 Pipeline build custom detector

C)LARA Traffic light detection data set

* The goal is : Detect traffic lights and classify them as red , Green and yellow

* Applications : It can be used for the intersection of road network adas Auto driving system rules

* Details : Three types of traffic 11K The frame and 20K + Annotation lights

* How to make use of data sets and establish the use of Mmdet-Faster-Rcnn-fpn50 Pipeline custom detection

D) Human detection using infrared image

* The goal is : Used to detect people in infrared images

* application : The autopilot car is equipped with infrared cameras to detect objects in bad conditions.

* Details :30 One with 1K + Annotated video sequence

* How to utilize data sets and use Mx-Rcnn Pipeline build custom detector

E) Data set of pothole detection

* The goal is : Detecting potholes from road images

* application : Smooth driving can be achieved by detecting road topography and potholes .

* Details : 700 One in the pit with 3K + Annotated images

* How to utilize data sets and use M-Rcnn Pipeline build custom detector

F)Nexet Vehicle detection data set

* The goal is : Detect road image of vehicle

* application : Vehicle detection is the main part of automatic driving

* Details :7000 Images , stay 6 There are 15K + notes

* How to utilize data sets and use Tensorflow Object Detection Building custom detectors API

G)BDD100K Adas Data sets

* The goal is : Detecting objects on the road

* application : Testing vehicles , Traffic signs and people are the main components of automatic driving

* Details :100K Images , Yes 10 Types of objects 250K + notes

* How to make use of data set and establish custom usage Tensorflow Object detection API The detector

H)Linkopings Traffic sign data set

* The goal is : Detection of traffic signs in images

* application : Detection of traffic signs is the first step to understand the traffic rules

* Details :3K Images , Yes 40 Various types of traffic signs are provided 5K + notes

** How to utilize data sets and use Mmdet-Cascade Mask-Rcnn Building custom detectors

fashion 、 Retail and marketing related data sets

A) Billboard detection ( Second sampling OpenImages Data sets ) Data sets

* The goal is : Detecting billboards in images

* Applications : The detection of billboards is a key part of automatic analysis of the whole city's marketing activities

* Details :2K Images , On the billboard 5K + notes

* How to utilize data sets and use Retinanet Building custom detectors

B)DeepFashion2 Fashion element detection data set

* The goal is : Detection of fashion products in images , clothes& decoration

* Applications : Application fashion detection has a huge application from data sorting to recommendation engine

* Details :490K Images , With about 100 Annotation object classes

* How to make use of data set and build customization CornetNet-Lite Pipeline detector

* Another data set related to fashion is Taobao commodity data set

C)Qmul-OpenLogo Logo detection dataset

* The goal is : Detect different logos in natural images

* application : Analyzing the frequency of logos in videos and natural scenes is crucial to marketing

* Details :16K Training pictures , Including the logo of various brands —— food 、 vehicle 、 Chain restaurant 、 Delivery Service 、 Airlines, etc

* How to utilize data sets and use mx-rcnn Pipeline build custom detector

Sports related data sets

A) Football detection data set ( from OpenImages The data set was resampled )

* The goal is : Cross frame detection of football in video

* application : The detection of football position is very important in the automatic analysis of offside

* Details : about 3K Training images .

* How to utilize data sets and use yolo-v3 Pipeline build custom detector

B) Card type detection

* The goal is : Detection of cards in natural images and classification of card types

* application : The possible application is to analyze the winning probability of different card games

* Details :52 In card types 500 More than images

* How to make use of data set and establish custom usage mx-rcnn Pipeline detector

C) Football player detection in thermal image

* The goal is : Using thermal images to locate and track players

* application : Tracking the players in the game is a key part of generation analysis

* Details : exceed 5K + The annotations 3K + Images .

* How to utilize data sets and use mmdet quick-rcnn Pipeline build custom detector

Security and monitoring related data sets

A)CCTV In traffic cameras MIO-TCD Vehicle detection

* The goal is : Detecting vehicles in CCTV cameras

* application : Detection of vehicles in CCTV cameras is a key part of security monitoring applications

* Details :113K Images , stay 5 On more than two types of vehicles 200K + notes

* How to utilize data sets and use Mmdet-Retinanet Pipeline build custom detector

B)WIDER Personnel detection data set

* The goal is : Detection of personnel in CCTV and natural scene images and videos

* application : be based on CCTV People detection is the core of security and monitoring applications

* Details :10K + Images and 20K + Notes can detect pedestrians

* How to make use of data set and establish custom usage Cornernet-Lite Pipe detector

C) Protective equipment - Helmet and vest inspection

* The goal is : Helmets and vests of inspectors

* application : This is an integral part of security compliance monitoring

* Details :1.5K + Images and 2K + Notes can detect people , Helmets and vests

* How to use data sets and build custom detectors Mmdet — Cascade RPN

D) Anomaly detection in video

* The goal is : The video is classified according to the operations performed in the video

* application : Real time detection of anomalies helps prevent crime

* Details : Corresponding to 10 Of exception categories 1K + video .

* How to utilize data sets and use mmaction-tsn50 Pipeline construction custom classifier

Medical image data set

A) Ultrasound brachial plexus (BP) Neural segmentation dataset

* The goal is : Segmentation of some nerve types in ultrasound images

* application : By using a indwelling catheter that can block or relieve pain at the source , Helps improve pain management .

* Details : 11K + Image and related example mask , Used to detect nerves

* How to use datasets and build custom detectors

B) In cells PanNuke Cancer case segmentation

* The goal is : Segmentation of different cell types in slide images

* Applications : The existence of cancer cells and dead cells in megabyte data

* Details :3K+ Images , With association instance mask for detecting different unit types

* How to use datasets and build custom detectors

Satellite imaging data set

A) Road segmentation in satellite image

* The goal is : Segmentation of trace routes in satellite images

* application : Help urban planning and road monitoring

* Details :1K + Image and related instance masks can detect different road regions

* How to use datasets and build custom detectors

B) Traversable region segmentation in synthetic lunar image

* The goal is : The rock is segmented and traversable areas are found in the lunar image

* application : Basic elements in path planning of autonomous Rover

* Details : With related instance mask 10K + Images to detect different rocks and flat ground

* How to utilize data sets and build custom detectors

C) Vehicle and swimming pool detection in satellite images

* The goal is : Vehicle and swimming pool detection in satellite images

* application : This is a key part of the property tax estimate

* Details :3.5K+ picture , There are cars and swimming pools 5K+ Comment tags

* How to utilize data sets and use cornernet lite Pipeline build custom detector

D) Road and residential area segmentation in aerial images

* The goal is : Segmentation of roads and residential areas in satellite images

* application : This is a key part of the property tax estimate

* Details : With split mask 100 Ultra high resolution images

* How to use datasets and build custom detectors

* Another similar road segmentation dataset and related training code

E) Water segmentation in satellite images

* The goal is : Segmentation of water body in satellite image

* application : It is important to understand how water bodies change and evolve over time

- With split mask 100 Ultra high resolution images

* How to use datasets and build custom detectors

* Another such dataset is DeepGlobe Land cover classification and its related use criteria

Wildlife related data sets

A) Tiger detection data set ( from OpenImages sampling )

* The goal is : Detect tigers in nature and UAV images

* application : Monitoring endangered species

* Details : with 4k + The annotations 2K + Images .

* How to utilize data sets and use Cornernet-lite Pipeline build custom detector

* Another such dataset could be a monkey detection dataset and related tutorials

B) Zebra and giraffe detection data set

* The goal is : Detection of zebra and giraffe species in natural and UAV images

* application : Monitoring endangered species

* Details : with 5k + The annotations 5K + Images .

* How to use datasets and use efficiencydet-d3 Pipeline build custom detector

C) Caltech camera trap dataset

* The goal is : Detect animals in trap camera type images

* application : Monitoring endangered species

* Details : with 8k + The annotations 10K + Images .

* How to utilize data sets and use Retinanet Pipeline build custom detector

* Another such camera dataset and related training code

D) Elephant detection data set ( from COCO Data set sampling )

* The goal is : Detection of elephant species in natural and UAV images

* application : Monitoring endangered species

* Details : with 5k + The annotations 5K + Images .

* How to utilize data sets and use mmdet-maskrcnn Building custom detectors

Underwater data set

A) Sea turtles found in the wild

* The goal is : Detecting turtles in underwater images

* application : Monitoring endangered species

* Details : with 5k + The annotations 5K + Images .

* How to use data set and use effective data to build custom detector

* Similar datasets , Monitoring of underwater fish

Related codes

B) Underwater waste detection data set

* The goal is : Detection of marine waste

* application : Monitoring and control of marine litter

* Details : with 5k + The annotations 2K + Images .

* How to use data set and use effective data to build custom detector

* More complex pixel based garbage classification data sets and related codes

C)SUIM Underwater object detection data set

* The goal is : Segmentation of underwater objects

* application : Path planning of autonomous underwater vehicle , Tracking divers and monitoring marine species

* Details :1.5K + Images and 1.5k + Notes mask .

* How to use datasets and build custom detectors

D) Salty underwater fish recognition data set

* The goal is : Detection of marine species in underwater images .

* Applications : Monitoring marine species

* Details :89 Videos to detect fish , Crab , shrimp , jellyfish , Starfish

* How to utilize data sets and use mmdet Building custom detectors ——Faster-rcnn The Conduit

Data sets related to text analysis

A) Document layout detection dataset

* The goal is : Examine document layout for further analysis

* application : It is necessary to segment the image into different parts , In order to further apply the rule-based NLP And text recognition .

* Details :5K + Images , with 10k + The label of the comment , Such as paragraph , Images , title .

* How to utilize data sets and use mx-rcnn Building custom detectors

* In a IIIT-AR-13K There are very similar datasets for graphic component detection in the document of , This is how to use data sets and train models on them

B) Total text dataset

* The goal is : Locating text in natural scenes

* Applications : Use OCR Basic components identified

* Details : with 5K + Polygon annotated 1.5K + Images

* How to utilize data sets and use Text-Snake Pipeline build custom detector

C)YY-Mnist Simple OCR Data sets

* The goal is : Locate and classify the numbers in the white background image

* Applications : Use OCR Basic components identified

* Details : exceed 10 Class has 2K + endorsed 1K Images

* How to utilize data sets and use Retinanet Pipeline build custom detector

Other datasets

A)TACO Garbage detection data set

* The goal is – Locating and segmenting all kinds of rubbish in image

* Applications : Key components of automated robots trying to solve the garbage problem in public places

* Details : contain 20 More than three types of garbage objects 15K + The annotations 10K Images

* How to utilize data sets and use Retinanet Pipeline build custom detector

B) General object detection data set for indoor scenes

* The goal is : Locating and detecting indoor objects in images

* Applications : Auto tag images in real estate and rental sites with Amenities

* Details : exceed 10 Different types of indoor objects ( For example, electrical appliances , The bed , The curtain , Chairs, etc )

* How to utilize data sets and use Retinanet Pipeline build custom detector

C)EgoHands Hand partial cut dataset

* The goal is : Segmentation of hands in natural scenes

* application : The first step in understanding gestures , And in human-computer interaction , Application of sign language recognition

* Details :4.8K + Image and corresponding hand mask .

* How to utilize data sets and use Retinanet Pipeline build custom detector

D)UCF Action recognition data set

* The goal is : The video is classified according to the operations performed in the video

* application : Marking video is very important for storing and retrieving a large number of videos

* Details : Corresponding to 101 Of action categories 1K + video .

* How to utilize data sets and use mmaction-tsn50 Pipeline construction custom classifier

E) Oil tank data set

* The goal is : Oil tank detection in satellite image

* application : Tracking tanks

* Details : have 10K + The annotations 10K + Images .

* How to utilize data sets and use Retinanet Pipeline construction custom classifier

Other action recognition data sets

A) Stair action recognition data set and how to train model on it

B)A2D Action recognition data set and how to train model on it

C)KTH Action recognition data set and how to train model on it

appendix

More details about the tutorial , Please visit our Github page

Link to the original text :https://medium.com/towards-ar...

Welcome to join us AI Blog station :
http://panchuang.net/

sklearn Machine learning Chinese official documents :
http://sklearn123.com/

Welcome to pay attention to pan Chuang blog resource summary station :
http://docs.panchuang.net/

版权声明
本文为[Artificial intelligence meets pioneer]所创,转载请带上原文链接,感谢