EfficientPose icon indicating copy to clipboard operation
EfficientPose copied to clipboard

color augmentation for custom dataset

Open jarvis-mark1 opened this issue 2 years ago • 2 comments

@ybkscht , I wanted to know what will be the size of dataset after applying the color augmentation if 'n' is given as 2 in randaug.py

example: I have created a dataset of 1000 images, what will be the number of images the model will get trained on if 'n' is given as 2 in randaug.py?

i went through your code and i thought the final dataset will be 1000 + 1000*2 = 3000 images

Am i correct?

jarvis-mark1 avatar Nov 03 '22 10:11 jarvis-mark1

hello! I want to ask some questions about the production of datasets, can you give me some help?

8umpk1n avatar Nov 12 '22 12:11 8umpk1n

Hi @jarvis-mark1, the size of the dataset will be the same as without augmentation because the color augmentation will be applied online during training. In RandAugment you got a list of possible augmentation techniques of which n will be sampled and applied to the image.

ybkscht avatar Jul 06 '23 12:07 ybkscht