def gen_test_data(test):
test_dir = '/content/test'
test_bad_dir = test_dir + '/bad'
test_good_dir = test_dir + '/good'
save = None
for idx in test:
tif = TIF(idx)
tif.gen_test(save=True, balance=True)
save = tif
return save.load_test_dir() # test data generator
lrs = [0.001, 0.0001, 0.00001]
decays = [1e-5, 1e-6, 1e-7]
for lr in lrs:
for decay in decays:
load_model_name=base+'/all-in-one-val'+'/model-epoch-'+str(i)+'-'+str(lr)+'-'+str(decay)
model = loadModel(load_model_name+'.json', load_model_name+'.h5')
sgd = optimizers.SGD(lr=lr, decay=decay, momentum=0.9, nesterov=True)
model.compile(optimizer=sgd,loss='binary_crossentropy',metrics=['accuracy'])
# AUC
prediction, prediction_score, gt = view_result_auc(test_dir, model)
y_pred = (np.transpose(prediction))
print(y_pred.shape)
print(y_pred)
y_true = gt
print(y_true.shape)
y_score = (np.transpose(prediction_score))
cm = confusion_matrix(y_true, y_pred)
np.save('cm-fin-'+str(pas)+'-epoch-'+str(i)+'-'+str(lr)+'-'+str(decay), cm)
plot_auc(y_true,y_score,'auc-fin-'+str(pas)+'-epoch-'+str(i)+'-'+str(lr)+'-'+str(decay)+'.png')
# heatmap
heat_prediction, heat_prediction_score = view_result_heatmap(TIF('078'), model)
plt.imsave('heatmap_prediction-fin-'+str(pas)+'.png',1-np.transpose(heat_prediction))
print('done')
#timestamp = str(int(time.time()))
timestamp = 'permute'
os.system('mkdir '+'permute')
base='/root/test2/'
number_of_pass = 10
model = initiate_model(new=True, load_model_name='')
epoch = 1
batch_size= 32
for pas in range(number_of_pass):
random.shuffle(train_val_id)
print("current pass---------", pas)
for tif_id in train_val_id:
tif = TIF(tif_id)
train(model,tif,epoch,batch_size)
model_name=base+timestamp+'/model-pass-'+str(pas+1)
# save model to file
saveModel(model, model_name)