'''
Run on GPU: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python use_charnet.py
'''
from __future__ import print_function
from numpy import argmax, array, dot, float32, mean, std, zeros
class CharNet():
def __init__(self, architecture_file=None, weight_file=None, optimizer=None):
# Generate mapping for softmax layer to characters
output_str = '0123456789abcdefghijklmnopqrstuvwxyz '
self.output = [x for x in output_str]
self.L = len(self.output)
# Load model and saved weights
from keras.models import model_from_json
if architecture_file is None:
self.model = model_from_json(open('char2_architecture.json').read())
else:
self.model = model_from_json(open(architecture_file).read())
if weight_file is None:
self.model.load_weights('char2_weights.h5')
else:
self.model.load_weights(weight_file)
if optimizer is None:
from keras.optimizers import SGD
optimizer = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
self.model.compile(loss='categorical_crossentropy', optimizer=optimizer)
def _rgb2gray(self,rgb):
return dot(rgb[...,:3], [0.2989, 0.5870, 0.1140])
def _preprocess(self,img):
from skimage.transform import resize
if len(img.shape) == 3 and img.shape[2] == 3:
img = self._rgb2gray(img)
img = resize(img, (32,100), order=1, preserve_range=True)
img = array(img,dtype=float32) # convert to single precision
img = (img - mean(img)) / ( (std(img) + 0.0001) )
return img
def classify_image(self,img):
img = self._preprocess(img)
xtest = zeros((1,1,32,100))
xtest[0,0,:,:] = img
z = self.model.predict(xtest,verbose=0)[0,:]
output_text = []
for i in range(23):
ind = argmax(z[i*self.L:(i+1)*self.L])
output_text.extend(self.output[ind])
return ''.join(output_text).rstrip(' ')
if __name__ == '__main__':
import matplotlib.image as mpimg
dir_prefix = '../IMAGES/'
filename = dir_prefix + 'Chevron.jpg'
#filename = dir_prefix + 'CondoleezzaRice.jpg'
#filename = dir_prefix + 'CMA_CGM.jpg'
cnn_model = CharNet()
img = mpimg.imread(filename)
print (cnn_model.classify_image(img))
