Skip to content

A complete Tensorflow implementation of cutout random erasing (without numpy)

Notifications You must be signed in to change notification settings

uranusx86/Random-Erasing-tensorflow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Random-Erasing-tensorflow

A data argumentation implementation of random erasing using Tensorflow
WITHOUT numpy !

Note: this side project is just for fun, the performance is equal to numpy version when using i5 CPU, but a bit lower than numpy version when using i7 CPU

Dependency

  1. Tensorflow 1.10+

Result

Test Performance

python3 test_performance.py

Current version test @ CPU i5-3470 & GPU GTX1080ti
numpy: 0.1695 sec/image
TF: 0.1696 sec/image

Ref

  1. Random Erasing Data Augmentation, Zhun Zhong, Liang Zheng, Guoliang Kang, Shaozi Li, Yi Yang
  2. Random Erasing (Author version)
  3. Cutout Random Erasing (Keras version)