# Image preprocessing
###### tags: `image preprocessing`


## 1. CNN Data (Classifying Label 1)
- /shared/data/image_cnn/
- Data Specs
- Downsample Detritus class 20,000
- Zooplankton class ~18,000
- Preprocessing:
- Gaussian blur kernel size 5
- Grayscale
- Resize to 40x40

## 2. Augmentation Data (Label 2)
- /shared/jenniferding/plankton
- Data Specs
- 500 samples from copepod and noncopepod data
- Preprocessing:
- Gaussian blur kernel size 3
- RGB
- Resize to 128x128

## Potential Issues:

There are 3 images that are labeled both copepod and noncopepod (label 2):
['Pia1.2017-10-26.1849+N00011891_hc.tif', 'Pia1.2017-10-03.1726+N00294605_hc.tif', 'Pia1.2017-10-03.1726+N00358839_hc.tif']

For now I've removed these from augmentation data