gradients = T.grad(cost, wrt=params)
gradient_acc = [theano.shared(0 * p.get_value()) for p in params]
counter = theano.shared(np.float32(0.))
acc = theano.function(inputs=[X, Y],
outputs=cost,
updates=[(a, a + g)
for a, g in zip(gradient_acc, gradients)] +
[(counter, counter + np.float32(1.))])
update = theano.function(
inputs=[],outputs=[],
updates = updates.momentum(params,[ g / counter for g in gradient_acc ]) \
+ [ (a, np.float32(0) * a) for a in gradient_acc ] \
+ [ (counter,np.float32(0.)) ]
)
test = theano.function(inputs=[X, Y], outputs=probs[:, Y])
training_examples = [word.strip() for word in open('dictionary.txt')]
import random
for _ in xrange(1500):
random.shuffle(training_examples)
for i, string in enumerate(training_examples):
print acc(font.imagify(string), label_seq(string))
if i % 20 == 0: update()
if i % 100 == 0:
hinton.plot(test(font.imagify("test"), label_seq("test")).T,
max_arr=1.)
hinton.plot(font.imagify("test").T[::-1].astype('float32'))
P.save('model.pkl')
import theano.tensor as T
import numpy as np
from theano_toolkit import utils as U
from theano_toolkit import hinton
from theano_toolkit import updates
from theano_toolkit.parameters import Parameters
import ctc
import font
import lstm
from ocr import *
if __name__ == "__main__":
import sys
test_word = sys.argv[1]
P = Parameters()
X = T.matrix('X')
predict = build_model(P, 8, 512, len(font.chars) + 1)
probs = predict(X)
test = theano.function(inputs=[X], outputs=probs)
P.load('model.pkl')
image = font.imagify(test_word)
hinton.plot(image.astype(np.float32).T[::-1])
y_seq = label_seq(test_word)
probs = test(image)
print " ", ' '.join(font.chars[i] if i < len(font.chars) else "_"
for i in np.argmax(probs, axis=1))
hinton.plot(probs[:, y_seq].T, max_arr=1.)
import theano.tensor as T
import numpy as np
from theano_toolkit import utils as U
from theano_toolkit import hinton
from theano_toolkit import updates
from theano_toolkit.parameters import Parameters
import ctc
import font
import lstm
from ocr import *
if __name__ == "__main__":
import sys
test_word = sys.argv[1]
P = Parameters()
X = T.matrix('X')
predict = build_model(P,8,512,len(font.chars)+1)
probs = predict(X)
test = theano.function(inputs=[X],outputs=probs)
P.load('model.pkl')
image = font.imagify(test_word)
hinton.plot(image.astype(np.float32).T[::-1])
y_seq = label_seq(test_word)
probs = test(image)
print " ", ' '.join(font.chars[i] if i < len(font.chars) else "_" for i in np.argmax(probs,axis=1))
hinton.plot(probs[:,y_seq].T,max_arr=1.)
acc = theano.function(
inputs=[X, Y],
outputs=cost,
updates = [
(a,a + g) for a,g in zip(gradient_acc,gradients)
] + [(counter,counter + np.float32(1.))]
)
update = theano.function(
inputs=[],outputs=[],
updates = updates.momentum(params,[ g / counter for g in gradient_acc ]) \
+ [ (a, np.float32(0) * a) for a in gradient_acc ] \
+ [ (counter,np.float32(0.)) ]
)
test = theano.function(
inputs=[X,Y],
outputs=probs[:,Y]
)
training_examples = [ word.strip() for word in open('dictionary.txt') ]
import random
for _ in xrange(1500):
random.shuffle(training_examples)
for i,string in enumerate(training_examples):
print acc(font.imagify(string),label_seq(string))
if i % 20 == 0: update()
if i % 100 == 0:
hinton.plot(test(font.imagify("test"),label_seq("test")).T,max_arr=1.)
hinton.plot(font.imagify("test").T[::-1].astype('float32'))
P.save('model.pkl')
predict = T.nnet.softmax(T.dot(hidden,W_hidden_output) + b_output)
return X,predict
def label_seq(string):
idxs = font.indexify(string)
result = np.ones((len(idxs)*2 + 1,),dtype=np.int32) * -1
result[np.arange(len(idxs))*2+1] = idxs
print result
return result
if __name__ == "__main__":
P = Parameters()
X = T.matrix('X')
Y = T.ivector('Y')
X,predict = build_model(P,X,8,256,len(font.chars)+1)
cost = ctc.cost(predict,Y)
params = P.values()
grad = T.grad(cost,wrt=params)
train = theano.function(
inputs=[X,Y],
outputs=cost,
updates=updates.adadelta(params,grad)
)
string = "Shawn"
print train(font.imagify(string),label_seq(string))
