g-bull/deep-phenotyping

Python codes for deep phenotyping using Theano and Keras

Deep-Phenotyping

Python codes for deep phenotyping using Theano and Keras

Our paper

Please reference our original paper when using these codes.

Sarah Taghavi Namin, Mohammad Esmaeilzadeh, Mohammad Najafi, Tim B. Brown, and Justin O. Borevitz, 'Deep Phenotyping: Deep Learning for Temporal Phenotype/Genotype Classification', bioRxiv, 2017.

Files description

cnn.py: Train and test with cnn (alexnet)

cnn_lstm.py: extract deep features using cnn, train and test lstm (label last frame)

cnn_lstm_perframe.py: extract deep features using cnn, train and test lstm (label each frame of sequence)

cnn_lstm_perframe_train.py: extract deep features using cnn, train lstm with all the data

cnn_lstm_perframe_test_rest_accessions.py: classify other accessions

cnn_featuremaps.py: feature maps form differnt layers of cnn

cnn_crf.py: extract deep features using cnn, train and test with CRF instead of lstm for temporal info

hcf_svm.py: handcrafted features, using svm for classification

hcf_lstm.py: handcrafted features, train and test with lstm

hcf_important_features.py: finding important handcrafted features

pots.py: preparing pots data

pots_with_area.py: preparing pots data + area

pots_with_more_features.py: preparing pots data + handcrafted features

pots_with_more_features_with_fourier.py: preparing pots data + handcrafted features + fourier features

Grabcut_segmentation_more_features_and_fourier.py: Segmentation and handcrafted features extraction

Dataset

https://figshare.com/s/e18a978267675059578f or http://phenocam.anu.edu.au/cloud/a_data/_webroot/published-data/2017/2017-Namin-et-al-DeepPheno.zip