def __init__(self, localizer=None, estimator=None):
'''
Parameters
----------
pixels: tuple #coordinates instead?
(img_rows, img_cols)
instrument: Instrument
Object resprenting the light-scattering instrument
model_path: str
path to model.h5 file
'''
if estimator is None:
self.estimator = Estimator()
else:
self.estimator = estimator
if localizer is None:
self.localizer = Localizer()
else:
self.localizer = localizer
self.instrument = estimator.instrument
if __name__ == '__main__':
from lmfit import report_fit
import cv2
import json
from matplotlib import pyplot as plt
keras_head_path = 'keras_models/predict_stamp_best'
keras_model_path = keras_head_path + '.h5'
keras_config_path = keras_head_path + '.json'
with open(keras_config_path, 'r') as f:
kconfig = json.load(f)
estimator = Estimator(model_path=keras_model_path, config_file=kconfig)
localizer = Localizer(configuration='holo', weights='_100k')
img_file = 'examples/test_image_large.png'
img = cv2.imread(img_file)
img_list = [img]
e2e = EndtoEnd(estimator=estimator, localizer=localizer)
features = e2e.predict(img_list=img_list)
example = features[0][0]
print('Example feature')
print(example.model.particle)
px = int(np.sqrt(example.data.size))
pix = (px, px)
cpix = estimator.pixels
class EndtoEnd(object):
'''
Attributes
__________
localizer: Localizer
Object resprenting the trained YOLO model
estimator: Estimator
Object representing the trained Keras model
instrument: Instrument
Object resprenting the light-scattering instrument
Methods
_______
predict(img_names_path=None, img_list=[], save_predictions=False, predictions_path='predictions.json', save_crops=False, crop_dir='./cropped_img')
loads img_names.txt from str 'img_names_path', imports images
img_names.txt contains absolute paths of images, separated by line break
or, just input images as a list
predicts on list of images using self.model
saves output to predictions_path if save_predictions = True
saves cropped images to crop_dir if save_crops = True
'''
def __init__(self, localizer=None, estimator=None):
'''
Parameters
----------
pixels: tuple #coordinates instead?
(img_rows, img_cols)
instrument: Instrument
Object resprenting the light-scattering instrument
model_path: str
path to model.h5 file
'''
if estimator is None:
self.estimator = Estimator()
else:
self.estimator = estimator
if localizer is None:
self.localizer = Localizer()
else:
self.localizer = localizer
self.instrument = estimator.instrument
@property
def coordinates(self):
return self._coordinates
@coordinates.setter
def coordinates(self, coordinates):
self._coordinates = coordinates
@property
def instrument(self):
return self._instrument
@instrument.setter
def instrument(self, instrument):
self._instrument = instrument
@property
def estimator(self):
return self._estimator
@estimator.setter
def estimator(self, estimator):
self._estimator = estimator
@property
def localizer(self):
return self._localizer
@localizer.setter
def localizer(self, localizer):
self._localizer = localizer
def predict(self, img_list=[], doubles_tol=0, edge_tol=0):
'''
output:
predictions: list of features
n images => n lists of features
'''
crop_px = self.estimator.pixels
yolo_predictions = self.localizer.predict(img_list=img_list)
yolo_predictions = nodoubles(yolo_predictions, tol=doubles_tol)
(imcols, imrows, channels) = img_list[0].shape
old_shape = (imrows, imcols)
yolo_predictions = no_edges(yolo_predictions,
tol=edge_tol,
image_shape=old_shape)
out_features, est_images, scales = crop_feature(
img_list=img_list, xy_preds=yolo_predictions, new_shape=crop_px)
structure = list(map(len, out_features))
char_predictions = self.estimator.predict(img_list=est_images,
scale_list=scales)
zpop = char_predictions['z_p']
apop = char_predictions['a_p']
npop = char_predictions['n_p']
for framenum in range(len(structure)):
listlen = structure[framenum]
frame = out_features[framenum]
index = 0
while listlen > index:
feature = frame[index]
feature.model.particle.z_p = zpop.pop(0)
feature.model.particle.a_p = apop.pop(0)
feature.model.particle.n_p = npop.pop(0)
feature.model.instrument = self.instrument
index += 1
return out_features
def localizer(self, localizer):
self._localizer = localizer or Localizer('tinyholo', weights='_500k')
self.sigLocalizerChanged.emit(self.localizer)