Author | Abhishek Annamraju compile | Flin source | medium

Computer vision is a rapidly evolving field, with a large number of new technologies and algorithms appearing in different conferences and journals every day. When it comes to target detection, in theory you’ll learn a lot of algorithms like faster-rCNN, Mask RCNN, Yolo, SSD, Retinenet, cascade RCNN, Peleenet, EfficientDet, CornerNet… . The list of algorithms is endless!

It’s always good to consolidate your learning experience by applying it to different data sets!!

As a result, you tend to have a better understanding of the algorithm and an intuitive understanding of which algorithms can run on which data sets.

Our open source team at Monk Computer Vision Org has compiled a list of object detection, image segmentation, and motion recognition datasets and created short tutorials for each object so you can work with these datasets and try out different object detection algorithms

Below is a short list of object detection datasets, brief details about them, and steps to use them. The data sets come from the following areas:

★ Agriculture ★ Advanced Driver Assistance and autonomous vehicle systems ★ Fashion, Retail and Marketing ★ Wildlife ★ Sports ★ satellite imaging ★ Medical imaging ★ Safety and surveillance ★ Underwater imaging

… . And more ！！！！！

The full list of instructions and training codes can be found on Github

  • Github.com/Tessellate-…

Data sets related to agriculture

A) Winegrape detects data sets

  • Github.com/thsant/wgis…

* Objective: To detect clusters of grapes in the vineyard

Applications: Monitor growth and analyze yield

* Details: 300 images with 4400 bounding boxes for 5 grape categories

* How to build a custom detector using a YoloV3 pipeline using a dataset

  • Github.com/Tessellate-…

B) Global wheat detection dataset

  • www.kaggle.com/c/global-wh…

* Objective: To test wheat crops in the field

Applications: Monitor growth and analyze yield

* Details: 3430 image with 100K + annotations

How to Build a custom detector with A Dataset and EfficientDet-D4 pipeline

  • Github.com/Tessellate-…

Advanced driver assistance and autonomous vehicle systems related data sets

A) LISA traffic sign detection data set

  • Cvrr.ucsd.edu/LISA/lisa-t…

* Objective: To detect and classify traffic signs in dashcam images

* Application: Traffic sign recognition is a rule setting procedure for autonomous driving

* Details: 7855 comments on 6610 frames on 47 US logo types

* How to utilize the data set and build a custom probe using Efficientdet-D3 pipeline

  • Github.com/Tessellate-…

* This repository has one more data set

  • LISA vehicle detection data
    • Cvrr.ucsd.edu/LISA/vehicl…

B) Object detection under low light conditions

  • Github.com/cs-chan/Exc…

* Objectives: Detect objects on the road in low light conditions — fog, haze, rain, etc

* Applications: This is an important part of self-driving cars because it is able to detect objects and therefore is a safer vehicle under adverse conditions

* Details: 15K + notes on 7500 frames on 12 different object types

How to Build a custom detector with a dataset and Efficientdet-D3 pipeline

  • Github.com/Tessellate-…

C) LARA traffic light detection data set

  • www.lara.prd.fr/benchmarks/…

* Goal: Detect traffic lights and classify them as red, green, and yellow

* Applications: Rules can be set for ADAS and autonomous vehicle systems at road network intersections

* Details: 11K frames and 20K + annotated lights for three traffic types

* How do I leverage the dataset and set up custom detections using the MMdet-ftP-RCNN-FPN50 pipeline

  • Github.com/Tessellate-…

D) Human detection using infrared images

  • camel.ece.gatech.edu/

* Target: Used to detect people in infrared images

* Applications: Self-driving cars are equipped with infrared cameras to detect objects in harsh conditions

* Details: 30 video sequences with 1K + annotations

* How do I leverage data sets and build custom detectors using the MX-RCNN pipeline

  • Github.com/Tessellate-…

E) Pothole detection data set

  • www.kaggle.com/chitholian/…

* Objective: Detect potholes from road images

* Application: Detect road terrain and potholes to achieve smooth driving.

* Details: 700 images with 3K + annotations in potholes

* How to build custom detectors using datasets and m-RCNN pipes

  • Github.com/Tessellate-…

F) Nexet vehicle detection data set

  • www.kaggle.com/solesensei/…

* Target: Detect road images of vehicles

* Applications: Detecting vehicles is a major part of autonomous driving

* Details: 7,000 images with 15K + annotations on 6 types of vehicles

* How to build a custom detector API using Tensorflow Object Detection using datasets

  • Github.com/Tessellate-…

G) BDD100K Adas dataset

  • www.kaggle.com/solesensei/…

* Targets: Detect objects on the road

* Applications: Detecting vehicles, traffic signs and people is a major part of autonomous driving

* Details: 100K images with 250K + annotations for 10 types of objects

* How do I leverage the dataset and build a custom detector that uses the Tensorflow object detection API

  • Github.com/Tessellate-…

H) Linkopings Traffic sign data set

  • www.cvl.isy.liu.se/research/da…

* Target: Detect traffic signs in the image

* Application: Checking traffic signs is the first step to understanding the rules of the road

* Details: 3K images, 5K + annotations for over 40 types of traffic signs

** How to build a custom detector using a dataset and mmDEt-Cascade mask-RCNN

  • Github.com/Tessellate-…

Fashion, retail and marketing related data sets

A) Billboard detection (secondary sampling OpenImages data set) data set

  • Storage.googleapis.com/openimages/…

* Target: Detect the billboard in the image

* Apps: Detecting billboards is a key part of automatically analyzing marketing campaigns throughout the city

* Details: 2K images with 5K + notes on the billboard

* How do I build custom detectors using Retinanet with datasets

  • Github.com/Tessellate-…

B) DeepFashion2 Fashion element detection dataset

  • Github.com/switchablen…

* Objectives: Detect images of fashion products, clothing and accessories

* Applications: Application fashion detection has huge applications ranging from data sorting to recommendation engines

* Details: 490K images with about 100 annotation object classes

* How to leverage data sets and build custom Cornetnet-Lite pipeline detectors

  • Github.com/Tessellate-…

* Another fashion-related data set is taobao merchandise data set

  • www.sysu-hcp.net/taobao-comm…

C) QMUL-OpenLOGO logo detection data set

  • qmul-openlogo.github.io/

* Goal: Detect different logos in natural images

* Application: Analyzing the frequency of logos in videos and natural scenes is critical to marketing

* Details: 16K training images, including logos for various brands — food, vehicles, chain restaurants, delivery services, airlines, etc

* How do I leverage data sets and build custom detectors using the MX-RCNN pipeline

  • Github.com/Tessellate-…

Sports related data sets

A) Football detection dataset (secondary sampling from OpenImages dataset)

  • Storage.googleapis.com/openimages/…

* Target: Detects the ball across frames in the video

Applications: Detection of football position is crucial in automatic analysis situations such as offside

* Details: approximately 3K training images.

* How to build a custom detector using the YOLO – V3 pipeline using a dataset

  • Github.com/Tessellate-…

B) Card type detection

  • www.kaggle.com/luantm/play…

* Goal: Detect cards in nature images and sort card types

Applications: Possible applications are analyzing the odds of winning different card games

* Details: 500 + images in 52 card types

* How to leverage data sets and build custom detectors using MX-RCNN pipes

  • www.kaggle.com/luantm/play…

C) Detection of football players in thermal images

  • www.kaggle.com/aalborguniv…

Goals: Use thermal graphics to locate and track the player

Apps: Tracking players in a game is a key part of generative analytics

* Details: 3K + images over 5K + notes.

* How do I build custom detectors using data sets and the MMDET Quick-RCNN pipeline

  • Github.com/Tessellate-…

Data sets related to security and monitoring

A) MiO-TCD vehicle detection in CCTV traffic cameras

* Objective: To detect vehicles in CCTV cameras

Applications: Detecting vehicles in CCTV cameras is a key part of security monitoring applications

* Details: 113K images with 200K + notes on 5 + types of vehicles

* How to build custom detectors using datasets and mmDet-Retinanet channels

  • Github.com/Tessellate-…

B) Data sets detected by WIDER personnel

  • wider-challenge.org/2019.html

* Targets: Detection of persons in CCTV and nature scene images and videos

* Applications: CCTV-based human detection forms the core of security and monitoring applications

* Details: pedestrians can be detected with 10K + images and 20K + notes

* How to leverage data sets and build custom detectors using the Cornernet-Lite pipeline

  • Github.com/Tessellate-…

C) Protective equipment – helmet and vest testing

  • Github.com/ciber-lab/p…

: : Objective: To detect personnel’s helmets and vests

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

* Details: 1.5K + images as well as 2K + notes can be detected for personnel, helmets and vests

* How to leverage datasets and build custom detectors using MMDET-Cascade RPN

  • Github.com/Tessellate-…

D) Anomaly detection in videos

  • www.crcv.ucf.edu/projects/re…

* Goals: Classify videos based on the actions performed in them

* Applications: Detecting anomalies in real time helps prevent crime

* Details: 1K + videos corresponding to 10 exception categories.

* How do I build custom classifiers with data sets and using the MMAction-TSN50 pipeline

  • Github.com/Tessellate-…

Medical image data set

A) Ultrasound brachial plexus (BP) nerve segmentation dataset

  • www.kaggle.com/c/ultrasoun…

* Objective: To segment certain neural types in ultrasound images

* Application: Helps improve pain management by using indwelling catheters that block or relieve pain at source.

* Details: 11K + images with associated instance masks for neural detection

* How do I leverage datasets and build custom detectors

  • Github.com/Tessellate-…

B) Segmentation of PanNuke cancer instances in cells

  • www.kaggle.com/andrewmvd/c…

* Objective: To segment different cell types in a slide image

* Applications: Automatically analyze megabytes of data for the presence of cancer and dead cells

* Details: 3K+ images with associated instance masks for detecting different cell types

* How to leverage data sets and build custom detectors

  • Github.com/Tessellate-…

Satellite imaging data set

A) Road segmentation in satellite images

  • www.kaggle.com/insaff/mass…

* Target: Split the path in the satellite image

* Applications: help with urban planning and road monitoring

* Details: 1K + images and associated instance masks can detect different road areas

* How do I leverage datasets and build custom detectors

  • Github.com/Tessellate-…

B) Segmentation of traversable regions in synthesized lunar images

  • www.kaggle.com/romainpessi…

* Goals: Split rocks and find traversable areas in the moon image

* Applications: Basic elements in path planning for autonomous roaming vehicles

* Details: 10K + images with related instance masks to detect different rocks and flat ground

* How do I leverage datasets and build custom detectors

C) Detection of cars and swimming pools in satellite imagery

  • www.kaggle.com/kbhartiya83…

* Objectives: Detect vehicles and swimming pools in satellite images

* Application: This is a key part of property tax estimation

* Details: 3.5K+ pictures, 5K+ annotated labels on cars and swimming pools

* How do I leverage datasets and build custom detectors using the Cornernet Lite pipeline

  • Github.com/Tessellate-…

D) Segmentation of roads and residential areas in aerial imagery

  • www.kaggle.com/cceekkigg/b…

* Objective: To segment roads and residential areas in satellite images

* Application: This is a key part of property tax estimation

* Details: 100 ultra high resolution images with segmentation masks

* How to leverage data sets and build custom detectors

  • Github.com/Tessellate-…

* Another similar road segmentation dataset and associated training code

  • Data set: www.kaggle.com/srikaranand…
  • Training code: github.com/Tessellate-…

E) Water segmentation in satellite images

  • www.kaggle.com/franciscoes…

* Objective: To segment water bodies in satellite images

* Applications: Is it important to understand how water bodies change and evolve over time

– 100 ultra high resolution images with segmentation mask

* How do I leverage datasets and build custom detectors

  • Github.com/Tessellate-…

* Another such data set is the DeepGlobe Land cover Classification and its associated guidelines for use

  • Data set: competitions.codalab.org/competition…
  • Guidelines for use: github.com/Tessellate-…

Wildlife-related data sets

A) Tiger detection dataset (sampled from OpenImages)

  • Storage.googleapis.com/openimages/…

* Targets: Detect tigers in nature and drone images

Applications: Monitoring endangered species

* Details: 2K + images with 4K + annotations.

* How do I build custom detectors using datasets and the Cornernet-Lite pipeline

  • Github.com/Tessellate-…

* Another such dataset could be the monkey detection dataset and its associated tutorial

  • The monkey test data set: storage.googleapis.com/openimages/…
  • Related tutorials: github.com/Tessellate-…

B) Zebra and giraffe detection data sets

  • Lev.cs.rpi.edu/public/data…

* Objectives: Detect zebra and giraffe species in nature and drone images

Applications: Monitoring endangered species

* Details: 5K + images with 5K + annotations.

* How to build a custom detector with a dataset and efficiencydet-D3 pipeline

  • Github.com/Tessellate-…

C) Caltech camera trap data set

  • Beerys. Making. IO/CaltechCame…

* Targets: Detect animals in trap camera type images

Applications: Monitoring endangered species

* Details: 10K + images with 8K + annotations.

* How do I build custom detectors using Retinanet channels with datasets

  • Github.com/Tessellate-…

* Another such camera data set and associated training code

  • Data set: github.com/Tessellate-…
  • Training code: github.com/Tessellate-…

D) Elephant detection dataset (sampled from COCO dataset)

  • cocodataset.org/#download

* Targets: Detect elephant species in nature and drone images

Applications: Monitoring endangered species

* Details: 5K + images with 5K + annotations.

* How to build custom detectors using MMDEt-MaskrCNN using data sets

  • Github.com/Tessellate-…

Underwater data set

A) Turtles found in the wild

  • Lev.cs.rpi.edu/public/data…

* Objective: Detect turtles in underwater images

Applications: Monitoring endangered species

* Details: 5K + images with 5K + annotations.

* How do I build custom detectors using datasets and valid data volumes

  • Github.com/Tessellate-…

* Similar data sets for monitoring underwater fish

  • Groups. Inf., Ed. The ac. UK/f4k/GROUNDT…

The relevant code

  • Github.com/Tessellate-…

B) Underwater garbage detection data set

  • Conservancy.umn.edu/handle/1129…

* Objective: To detect Marine debris

Applications: Monitoring and controlling Marine debris problems

* Details: 2K + images with 5K + annotations.

* How do I build custom detectors using datasets and valid data volumes

  • Github.com/Tessellate-…

* More complex pixel-based garbage sorting data sets and associated code

  • Garbage sorting data sets: conservancy.umn.edu/handle/1129…
  • Related code: github.com/Tessellate-…

C) SUIM underwater object detection data set

  • Irvlab.cs.umn.edu/resources/s…

* Objective: Split underwater objects

* Applications: path planning for autonomous underwater vehicles, tracking divers and monitoring Marine species

* Details: 1.5K + images and 1.5K + comment masks.

* How do I leverage datasets and build custom detectors

  • Irvlab.cs.umn.edu/resources/s…

D) Brackish underwater fish identification data sets

  • www.kaggle.com/aalborguniv…

* Objective: To detect Marine species in underwater images.

* Apps: Monitor Marine species

* Details: 89 videos to detect fish, crabs, shrimp, jellyfish, starfish

* How do I build a custom detector — ftP-RCNN pipeline using a dataset and MMDET

  • Github.com/Tessellate-…

Data sets related to text analysis

A) Document layout detection data set

  • www.primaresearch.org/datasets/La…

* Goal: Detect document layout for further analysis

* Application: It is essential to divide the image into different parts so that rules-based NLP and text recognition functions can be further applied.

* Details: 5K + images, labels with 10K + annotations, such as paragraphs, images, headings.

* How do I leverage data sets and build custom detectors using MX-RCNN

  • Github.com/Tessellate-…

* A very similar data set exists for graphical component detection in a document named IIIT-AR-13K. How is this a way to take advantage of the data set and train the model on it

  • Github.com/Tessellate-…

B) Total textual data set

  • Github.com/cs-chan/Tot…

* Target: Position text in a natural scene

* Applications: Basic components identified using OCR

* Details: 1.5K + image with 5K + polygon annotation

* How to build custom detectors using data sets and the Text-Snake pipeline

  • Github.com/Tessellate-…

C) YY-MNIST simple OCR dataset

* Target: Locate and classify numbers in a white background image

* Applications: Basic components identified using OCR

* Details: more than 10 categories of 1K images with 2K + annotations

* How do I build custom detectors using Retinanet channels with datasets

  • Github.com/Tessellate-…

Other data sets

A) TACO garbage detection data set

  • tacodataset.org/

* Target – Locate and segment various garbage in the image

* Apps: Key components of autonomous robots that attempt to solve garbage problems in public places

* Details: 15K + annotated 10K images containing more than 20 different categories of garbage objects

* How do I build custom detectors using Retinanet channels with datasets

  • Github.com/Tessellate-…

B) Universal object detection data set for indoor scenes

  • Storage.googleapis.com/openimages/…

* Target: Locate and detect indoor objects in the image

* Apps: Automatically tag images on real estate and rental sites with amenities

* Details: More than 10 different categories of interior objects (e.g. appliances, beds, curtains, chairs, etc.)

* How to build custom detectors using datasets and Retinanet channels

  • Github.com/Tessellate-…

C) EgoHands hand segmentation dataset

  • Vision.soic.indiana.edu/projects/eg…

* Goal: Split hands in a natural scene

* Application: The first step in understanding gestures, and applications in human-computer interaction, sign language recognition

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

* How to build custom detectors using datasets and Retinanet channels

  • Github.com/Tessellate-…

D) UCF action recognition data set

  • www.crcv.ucf.edu/data/UCF101…

* Goals: Classify videos based on the actions performed in them

* Applications: Tagging video is important for storing and retrieving large amounts of video

* Details: 1K + videos for 101 action categories.

* How do I build custom classifiers with data sets and using the MMAction-TSN50 pipeline

  • Github.com/Tessellate-…

E) Tank data set

  • www.kaggle.com/towardsentr…

* Objective: To detect oil tanks in satellite images

* Applications: Tracking oil tanks

* Details: 10K + images with 10K + annotations.

* How do I build a custom classifier using a dataset and a Retinanet pipeline

  • Github.com/Tessellate-…

Other action recognition data sets

A) Stair movement recognition data set and how to train models on it

  • Data set: the actions. The stair. Center/videos. HTML

  • Training model: github.com/Tessellate-…

B) A2D action recognition dataset and how to train models on it

  • Data set: web.eecs.umich.edu/~jjcorso/r/…

  • Training model: github.com/Tessellate-…

C) THE KTH action recognition dataset and how to train models on it

  • Data set: www.csc.kth.se/cvap/action…

  • Training model: github.com/Tessellate-…

The appendix

For more details on the tutorial, visit our Github page

  • Github.com/Tessellate-…

Original link: medium.com/towards-art…

Welcome to panchuangai blog: panchuang.net/

Sklearn123.com/

Welcome to docs.panchuang.net/