Skip to content

Open Images is a dataset of ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, visual relationships, and localized narratives. Download Your own dataset by entering the given commands in CMD or Anaconda Prompt

License

Notifications You must be signed in to change notification settings

TusharKalyankar/OIDv4_Toolkit-Custom-YOLOv3-Dataset-Collector

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Open Images is a dataset of ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, visual relationships, and localized narratives.

1.0 Download Your own dataset by entering the given commands in CMD or Anaconda Prompt

Step1)-- Check for your convinient dataset on https://storage.googleapis.com/openimages/web/visualizer/index.html?set=train&type=detection&c=%2Fm%2F044r5d

Step2)-- Go to--> Type--> Select Detection

Step3)-- Go to--> Options --> Deselect-- Display Boxes from all Catagory

Step4)-- Make Sure to input the same class you Searched after "--classes" in the COMMAND BELOW

  -- There is e.g. in COOMAND BELOW for class:  "Frying Pan" and "Refrigerator"

Step5)-- Mention the command --limit and the number of images for each class

Step6)-- Modify Below command accordingly and Run in the directory of "OIDV4_Toolkit" where "main.py" is present

Step7)-- Type "Y" If asked for the permission

Run the following command

  -- python main.py downloader --classes Frying_Pan Refrigerator --type_csv train --limit 200 --multiclasses 1 

Step8)-- After the download your dataset of Yolo wiil be present in OIDv4_Toolkit-Custom-Dataset-Collector-->OID-->Dataset-->train

Step9)-- In the downloaded repository you will get "classes.txt" Modify that with your own classes //ONE CLASS PER LINE

Step10)-- Run python "convert_annotations.py"

Your CUSTOM YOLO DATASET IS READY

Reference

"We don't need no bounding-boxes: Training object class detectors using only human verification"Papadopolous et al., CVPR 2016.

Refered from the video on Youtube Reference: https://www.youtube.com/watch?v=_4A9inxGqRM

About

Open Images is a dataset of ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, visual relationships, and localized narratives. Download Your own dataset by entering the given commands in CMD or Anaconda Prompt

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages