woolimi/leaffliction

leaffliction

An innovative computer vision project utilizing leaf image analysis for disease recognition.

How to use

Download dataset

# Download image dataset and generate distribution chart image
python 01.Distribution.py apple grape

Data augmentation

# Augment unbalanced image dataset
python 02.Augmentation.py

Auto image augmentation...

Augmenting "Apple" images...
========================
Summary of augmentation:
========================
/Users/woolim/Documents/leaffliction/images/Apple_scab: 629 -> 1640
/Users/woolim/Documents/leaffliction/images/Apple_healthy: 1640 -> 1640
/Users/woolim/Documents/leaffliction/images/Apple_rust: 275 -> 1640
/Users/woolim/Documents/leaffliction/images/Apple_Black_rot: 620 -> 1640

Augmenting "Grape" images...
========================
Summary of augmentation:
========================
/Users/woolim/Documents/leaffliction/images/Grape_Esca: 1382 -> 1382
/Users/woolim/Documents/leaffliction/images/Grape_healthy: 422 -> 1382
/Users/woolim/Documents/leaffliction/images/Grape_Black_rot: 1178 -> 1382
/Users/woolim/Documents/leaffliction/images/Grape_spot: 1075 -> 1382

Check image distributions are well balanced

python 01.Distribution.py apple grape

apple before	apple after

grape before	grape after

Save transformed image plot

python 03.Transformation.py -src [SRC_PATH] -dst [DST_PATH]

image transformed	image transformed

Predict an image

python predict.py [image_path]

Predict all images and check prediction accuracy.

python 04.Classification.py

Validation Progress: 100%|███████████████████████████████████████████| 10/10 [01:10<00:00,  7.01s/it]
Accuracy of the model on the validation set: 92.31%

Tensorboard

tensorboard --logdir runs

Train loss	Train vs Validation Loss

Resources

• Youtube Coursera CNN