def runModel(self):
self.text = ""
letters, _ = get_letters(get_processed_image(self.image))
for letter in letters:
self.text += get_char(self.model.predict(letter).argmax())
print(self.text)
def predict(image_path, model, top_k, category_names, dev=False):
if dev:
print(image_path, model, top_k, category_names, dev)
image = get_processed_image(image_path)
model = load_model(model)
class_names = get_class_names(category_names)
prediction = model.predict(np.expand_dims(image, axis=0))
values, indices = tf.math.top_k(prediction, top_k)
values = values.numpy()[0]
classes = [class_names[str(value+1)] for value in indices.cpu().numpy()[0]]
if dev:
print("model.summary")
print(model.summary())
print(f'top class: {classes[0]} with % {values[0]*100}')
print(f'values: {values}\nclasses: {classes}')
return values, classes
def save(self, name, content, save=True):
if self and not self._committed and None in self.patterns:
# Apply original image process
img = self._get_image()
processed = get_processed_image(self, img, self.patterns[None])
file_fmt = get_fileformat_from_filename(name)
content = get_content_file(processed, file_fmt)
super(ThumbnailFieldFile, self).save(name, content, save=save)
def _create_thumbnail(self, patterns):
"""create PIL thumbnail of this field file
Attribute:
patterns -- A process patterns to generate thumbnail
"""
img = self._get_image()
thumb = get_processed_image(self, img, patterns)
return thumb
