creating image dataset | Coursera Community
Coursera Header
Solved

creating image dataset

  • 18 October 2019
  • 7 replies
  • 129 views

Badge +1

Can anyone guide me on creating image dataset from scratch.…

I have started working on a new project but no online dataset available. So i am clicking pictures to make my own dataset. But dont know how to prepare them for feeding the model.

Any rough help or guidance will be appreciated

icon

Best answer by Kalyan 20 October 2019, 08:28

Hey @Ananda Basak 

In general you can follow the below points for creating image data set from scratch :

  1. Download a set of images from somewhere.
  2. Make sure they have the same extension (.jpg or .png for instance)
  3. Make sure that they are named according to the convention of the first notebook i.e. class.number.extension for instance cat.14.jpg)
  4. Split them in different subsets like train, valid, and test.

I hope this helps.

Cheers

View original

7 replies

Badge

@Ananda Basak 

What type of model do you plan to use and which library do you plan to use? That will help you decide what format the image date needs to be in to feed the model.

Userlevel 4
Badge +4

Hey @Ananda Basak 

In general you can follow the below points for creating image data set from scratch :

  1. Download a set of images from somewhere.
  2. Make sure they have the same extension (.jpg or .png for instance)
  3. Make sure that they are named according to the convention of the first notebook i.e. class.number.extension for instance cat.14.jpg)
  4. Split them in different subsets like train, valid, and test.

I hope this helps.

Cheers

Badge

Hey there @Ananda Basak 

You can create an image classifier with nothing but keywords.

Check my Notebook attached.

Image classifier with nothing but keywords

Badge +1

Ok thanks @hhaider5 @Kalyan 

Badge

Ok thanks @hhaider5 @Kalyan 

Anytime.

Badge

Hey there,

If you have limited amount of images , then just try using OpenCV library, perform Image augmentation like rotate the image, re scale those, try to use gaussian blur after re scaling the images.

you can just run for loop for above process  

Badge

You could inspire yourself with previously build dataset like the PlantVillage dataset (https://github.com/spMohanty/PlantVillage-Dataset). 

Reply