def train(ftrain, ftest, epochs, savemodel, saveloss, savetest):
X, y = load2d(ftrain, test=False)
net3 = create_network(epochs)
net3.fit(X, y)
sys.setrecursionlimit(1500000)
with open(savemodel, 'wb') as f:
pickle.dump(net3, f, -1)
draw_loss_2(net3, saveloss)
test(net3, ftest, savetest)
def fine_tune(fmodel, ftrain, epochs, ftest, savemodel, saveloss, savetest):
X1, y1 = load2d(ftrain, test=False)
listFrozens = [] #['conv1','conv2','conv3']
newlayers = set_weights(fmodel, frozen=False, listLayers=listFrozens)
net2 = build_model1(newlayers, epochs)
net2.fit(X1, y1)
sys.setrecursionlimit(1500000)
with open(savemodel, 'wb') as f:
pickle.dump(net2, f, -1)
draw_loss_2(net2, saveloss)
test(net2, ftest, savetest)
def test(net, ftest, fsave):
X, _ = load2d(ftest, test=True)
y_pred = net.predict(X)
fig = pyplot.figure(figsize=(18, 16))
fig.subplots_adjust(left=0,
right=1,
bottom=0,
top=1,
hspace=0.05,
wspace=0.05)
for i in range(16):
ax = fig.add_subplot(4, 4, i + 1, xticks=[], yticks=[])
plot_sample(X[i], y_pred[i], ax)
fig.savefig(fsave, dpi=90)
pyplot.close(fig)
FSAVEFOLDER = '/data3/linhlv/OUTPUT2/2018/tete/fine_tuning/run_test_10landmarks/'
filename = FSAVEFOLDER + 'landmarks/cnnmodel_10_output_fine_tuning_unfreeze_' #.txt
FSAVEIMAGES = FSAVEFOLDER + 'images/'
DATA = ['v10', 'v11', 'v12', 'v14', 'v15', 'v16', 'v17', 'v18', 'v19']
for i in DATA:
fmodelf = FMODEL + i + '.pickle'
ftestf = FTEST + i + '.csv'
flandmarks = filename + i + '.txt'
net = None
sys.setrecursionlimit(100000)
with open(fmodelf, 'rb') as f:
net = pickle.load(f)
X, _ = load2d(ftestf, test=True)
y_pred = net.predict(X)
# try to display the estimated landmarks on images
paths = loadCSV(ftestf)
fileNames = extract_fileNames(paths)
for i in range(len(y_pred)):
predi = y_pred[i]
#filename = FSAVEFOLDER + FNAMES[i]
write_file(flandmarks, predi)
#write_file(filename,"\n")
saveImg = FSAVEIMAGES + fileNames[i]
fig = pyplot.figure()
ax = fig.add_subplot(1, 1, 1, xticks=[], yticks=[])
plot_sample(X[i], predi, ax)
print(model)
all_param = lasagne.layers.get_all_param_values(model.layers)
net = build_model()
lasagne.layers.set_all_param_values(net['output'],all_param,trainable=True)
newlayers = lasagne.layers.DenseLayer(net['hidden5'],num_units = 16, nonlinearity=None)
#model.layers = newlayers
print(model)
return model
'''
if __name__ == '__main__':
# Load data
FTRAINF = '/data3/linhlv/pronotum/v1/csv/train_v19.csv'
FTESTF = '/data3/linhlv/pronotum/v1/csv/test_v19.csv'
X1,y1 = load2d(FTRAINF,test=False)
#=================================================================
# Load the parameters into list of layer, create a new network and train
newlayers = set_weights(FMODEL)
net2 = build_fine_tuning_model(newlayers)
net2.fit(X1,y1)
# Save the fine-tuning model
sys.setrecursionlimit(150000)
with open('/data3/linhlv/2018/saveModels/cnnmodel_all_10000_pronotum_fine_tune_v19.pickle','wb') as f:
pickle.dump(net2,f,-1)
# draw the loss
draw_loss(net2)