def create_new_shrinking_factor_model(input_path, new_shrinking_factor, output_path):
input_model = read_model(input_path)
model_class = input_model.__class__
output_model = input_model
if True:
if model_class is MultiScalesDetectorModel:
for detector in output_model.detectors:
change_shrinking_factor(new_shrinking_factor, detector)
elif model_class is DetectorModel:
change_shrinking_factor(new_shrinking_factor, output_model)
else:
raise Exception("Received an unmanaged detector model class {0}".format(model_class) )
#replace_scale_hack(output_model)
output_content = output_model.SerializeToString()
out_file = open(output_path, "wb")
out_file.write(output_content)
out_file.close()
print("Created output model file", output_path)
return
def generate_occlusionClassifier(input_model):
width = 32
half_width = 16
model = read_model(input_model)
for i in range(1 + half_width):
yThresh = half_width - i
new_model = DetectorModel()
new_model.CopyFrom(model)
if model.model_window_size:
model_width = model.model_window_size.x
model_height = model.model_window_size.y
print("model.detector_type", model.detector_type)
if model.detector_type == model.SoftCascadeOverIntegralChannels:
old_cascade = model.soft_cascade_model
new_cascade = new_model.soft_cascade_model
print("Model has %i stages" % len(old_cascade.stages))
update_cascade(old_cascade, new_cascade, width - yThresh)
output_path = input_model + "_artificial_crop_" + str(yThresh * 4)
out_file = open(output_path, "wb")
out_file.write(new_model.SerializeToString())
out_file.close()
print("Create output model file", output_path)
def generate_occlusionClassifier(input_model):
width = 32
half_width = 16
model=read_model(input_model)
for i in range(1+half_width):
yThresh = half_width-i
new_model = DetectorModel()
new_model.CopyFrom(model)
if model.model_window_size:
model_width = model.model_window_size.x
model_height = model.model_window_size.y
print("model.detector_type", model.detector_type)
if model.detector_type == model.SoftCascadeOverIntegralChannels:
old_cascade = model.soft_cascade_model
new_cascade = new_model.soft_cascade_model
print("Model has %i stages" % len(old_cascade.stages))
update_cascade(old_cascade, new_cascade, width-yThresh)
output_path = input_model + "_artificial_crop_" + str(yThresh*4)
out_file = open(output_path, "wb")
out_file.write(new_model.SerializeToString())
out_file.close()
print("Create output model file", output_path)
def main():
#options = parse_input_options()
#model_filename = options.input_path
boosted_model_path = "../../src/applications/boosted_learning/2012_04_16_74669_full_size_model_non_recursive_seed_22222_push_up_0.05.bootstrap2crop_32"
svm_model_path = "../../libs/liblinear-1.8/tmp.model"
output_boosted_model_path = "test_model_c0.001_e0.001_dataBalanced.out.proto.bin"
assert boosted_model_path != output_boosted_model_path
print("Reading boosted model", boosted_model_path)
model = read_model(boosted_model_path)
print("Reading svm model", svm_model_path)
w_vector = get_svm_model_w_vector(open(svm_model_path, "r"))
if type(model) is detector_model_pb2.DetectorModel:
print("Cutting model...")
cut_model(model, len(w_vector))
print("Saving cut model at", output_boosted_model_path + ".cut")
out_file = open(output_boosted_model_path + ".cut", "wb")
out_file.write(model.SerializeToString())
out_file.close()
print("Adjusting weights model...")
adjust_weights(model, w_vector)
print("Saving result model at", output_boosted_model_path)
out_file = open(output_boosted_model_path, "wb")
out_file.write(model.SerializeToString())
out_file.close()
output_boosted_model_path
print("Created output model file", output_boosted_model_path)
else:
raise Exception("No model type other than DetectorModel is supported. "
"Received %s" % type(model))
pylab.show() # blocking call
return
def main():
#options = parse_input_options()
#model_filename = options.input_path
boosted_model_path = "../../src/applications/boosted_learning/2012_04_16_74669_full_size_model_non_recursive_seed_22222_push_up_0.05.bootstrap2crop_32"
svm_model_path = "../../libs/liblinear-1.8/tmp.model"
output_boosted_model_path = "test_model_c0.001_e0.001_dataBalanced.out.proto.bin"
assert boosted_model_path != output_boosted_model_path
print("Reading boosted model", boosted_model_path)
model = read_model(boosted_model_path)
print("Reading svm model", svm_model_path)
w_vector = get_svm_model_w_vector(open(svm_model_path, "r"))
if type(model) is detector_model_pb2.DetectorModel:
print("Cutting model...")
cut_model(model, len(w_vector))
print("Saving cut model at", output_boosted_model_path + ".cut")
out_file = open(output_boosted_model_path + ".cut", "wb")
out_file.write(model.SerializeToString())
out_file.close()
print("Adjusting weights model...")
adjust_weights(model, w_vector)
print("Saving result model at", output_boosted_model_path)
out_file = open(output_boosted_model_path, "wb")
out_file.write(model.SerializeToString())
out_file.close()
output_boosted_model_path
print("Created output model file", output_boosted_model_path)
else:
raise Exception("No model type other than DetectorModel is supported. "
"Received %s" % type(model) )
pylab.show() # blocking call
return
def replace_scale_hack(model):
"""
Small hack to replace a model of a given scale
"""
assert isinstance(model, MultiScalesDetectorModel)
model_path = "/users/visics/mmathias/devel/doppia/src/applications/boosted_learning/eccvWorkshop/2012_05_20_67022_trained_model_octave_-1.proto.bin.bootstrap2"
replacement_model = read_model(model_path)
assert isinstance(replacement_model, DetectorModel)
replacement_scale = 0.5
print("Replacing model of scale {0} " \
"by the model read at {1}".format(replacement_scale, model_path))
for detector in model.detectors:
if detector.scale == replacement_scale:
detector.CopyFrom(replacement_model)
detector.scale = replacement_scale # we set the proper new scale
break
return
if not is_the_same:
print("A"); read_tree(a[0], a[1])
print("B"); read_tree(a[0], a[1])
print("Found two different stages after %i comparisons" % counter)
differences += 1
if differences > 5:
raise Exception("To many differences, the two models are not identical")
# end of compare_detectors
print("Reading the two models; please wait...")
#model_a_path = "../../src/applications/boosted_learning/2011_10_12_64789_trained_model_octave_0.proto.softcascade.bin"
#model_b_path = "../../src/applications/boosted_learning/2011_11_04_61124_2011_10_12_64789_trained_model_octave_0.proto.softcascade.softcascade.bin"
model_a_path = "/users/visics/rbenenso/code/europa/code/kul-rwth/objects_detection_lib/2012_04_04_1417_trained_model_multiscales_synthetic_softcascade.proto.bin"
model_b_path = "2012_04_04_1417_trained_model_multiscales_synthetic_softcascade.shrinking_factor_2_back_to_4.proto.bin"
model_a = read_model(model_a_path)
model_b = read_model(model_b_path)
if isinstance(model_a, detector_model_pb2.MultiScalesDetectorModel):
assert isinstance(model_b, detector_model_pb2.MultiScalesDetectorModel)
for i in range(len(model_a.detectors)):
detector_a = model_a.detectors[i]
detector_b = model_b.detectors[i]
compare_detectors(detector_a, detector_b)
else:
compare_detectors(model_a, model_b)
print("End of game, have a nice day !")
read_tree(a[0], a[1])
print("Found two different stages after %i comparisons" % counter)
differences += 1
if differences > 5:
raise Exception(
"To many differences, the two models are not identical")
# end of compare_detectors
print("Reading the two models; please wait...")
#model_a_path = "../../src/applications/boosted_learning/2011_10_12_64789_trained_model_octave_0.proto.softcascade.bin"
#model_b_path = "../../src/applications/boosted_learning/2011_11_04_61124_2011_10_12_64789_trained_model_octave_0.proto.softcascade.softcascade.bin"
model_a_path = "/users/visics/rbenenso/code/europa/code/kul-rwth/objects_detection_lib/2012_04_04_1417_trained_model_multiscales_synthetic_softcascade.proto.bin"
model_b_path = "2012_04_04_1417_trained_model_multiscales_synthetic_softcascade.shrinking_factor_2_back_to_4.proto.bin"
model_a = read_model(model_a_path)
model_b = read_model(model_b_path)
if isinstance(model_a, detector_model_pb2.MultiScalesDetectorModel):
assert isinstance(model_b, detector_model_pb2.MultiScalesDetectorModel)
for i in range(len(model_a.detectors)):
detector_a = model_a.detectors[i]
detector_b = model_b.detectors[i]
compare_detectors(detector_a, detector_b)
else:
compare_detectors(model_a, model_b)
print("End of game, have a nice day !")
