LeavesRecognitionCNN

This is an implementation of paper [A Convolutional Neural Network for Leaves Recognition Using Data Augmentation] (https://ieeexplore.ieee.org/document/7363364).

Prerequisites

• Python 3.6
• keras 2.2.4 *script written below is for UBUNTU but you can run it on windows as well.

Running

for training
python main-run.py --numEpochs 100 \
--imgSize (256,256) \
--momentum 0.9 \
--decay 0.06 \
--learnRate 0.01 \
--batchSize 80 \
--noOfLayers 5\
--lossfn 'categorical_crossentropy'\
--outDir 'outData'\
--inpDir './data'\
--loadModel None \
--plot True\
--noOfWorkers 4\
--dataAug True


This saves the trained model in specified output directory and plots different training metrics.

for testing
python main-run.py --numEpochs 100 \
--imgSize (256,256) \
--momentum 0.9 \
--decay 0.06 \
--learnRate 0.01 \
--batchSize 80 \
--noOfLayers 5\
--lossfn 'categorical_crossentropy'\
--outDir 'outData'\
--inpDir './data'\
--loadModel 'Datamodel.hdf5'\
--plot True\
--noOfWorkers 4\
--dataAug True

This loads the specified model and prints test loss and accuracy.