https://github.com/chekoduadarsh/random-pixel-drop

https://github.com/chekoduadarsh/random-pixel-drop

augmentation-libraries dropout dropped-pixels image-augmentation-preprocessing imagedatagenerator

Last synced: 19 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/chekoduadarsh/random-pixel-drop
• Owner: chekoduadarsh
• Created: 2020-02-05T13:52:26.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-02-13T18:27:03.000Z (almost 5 years ago)
• Last Synced: 2024-10-28T14:38:02.648Z (2 months ago)
• Topics: augmentation-libraries, dropout, dropped-pixels, image-augmentation-preprocessing, imagedatagenerator
• Language: Python
• Size: 11.7 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Random Pixel Drop

This is a library written for `ImageDataGenerator()` from `keras.preprocessing` library.

Random pixel drop is inspired from dropout regularisation technique and cutout image augmentation technique. This library will randomly set image pixel values of the image to zeros for a given range of number. Which will be similar to dropout than switching neurons randomly to reduce over fitting this will introduce randomness in the input image to reduce over fitting 

# Implementation

```

git clone https://github.com/chekoduadarsh/Random-Pixel-Drop

```

The above given command will download the repository to your working directory, then in your python code tou can import the library as following,

## With ImageDataGenerator in Keras

It is very easy to use if you are using ImageDataGenerator in Keras, get `drop` function from `get_random_pixel_drop()` which can be passed to `ImageDataGenerator` as `preprocessing_function`. This function uses 3 hyper parameters

`Drop_num` -> Number of dropped pixels

`l_rand_bound` -> lower boundry for random pixel value generation

`u_rand_bound` -> upper boundry for random pixel value generation