def bolt_obj_2_feature_vector(all_features_obj_dict, feature_name_list):
"""
For each object - pull out features and store in feature_obj
with the same structure as all_bolt_data
Dictionary - "tap", "slide", "slow_slide",
"thermal_hold", "squeeze"
Directly store the features into a vector
See createFeatureVector for more details on structure
"""
# Store in feature class object
all_features_vector_dict = dict()
# For all motions
for motion_name in all_features_obj_dict:
feature_obj_list = all_features_obj_dict.get(motion_name)
all_adjective_labels_dict = dict()
feature_vector_list = list()
feature_label_list = list()
# For all objects
for bolt_feature_obj in feature_obj_list:
# Create feature vector
feature_vector, object_id = utilities.createFeatureVector(
bolt_feature_obj, feature_name_list)
feature_vector_list.append(feature_vector)
feature_label_list.append(object_id)
# Create label dictionary
labels = bolt_feature_obj.labels
for adjective in labels:
# Check if it is the first time adjective added
if (all_adjective_labels_dict.has_key(adjective)):
adjective_array = all_adjective_labels_dict[adjective]
else:
adjective_array = list()
# Store array
adjective_array.append(labels[adjective])
all_adjective_labels_dict[adjective] = adjective_array
# Store all of the objects away
all_features_vector_dict[motion_name] = (np.array(feature_vector_list),
feature_label_list)
return (all_features_vector_dict, all_adjective_labels_dict)
def bolt_obj_2_feature_vector(all_features_obj_dict, feature_name_list):
"""
For each object - pull out features and store in feature_obj
with the same structure as all_bolt_data
Dictionary - "tap", "slide", "slow_slide",
"thermal_hold", "squeeze"
Directly store the features into a vector
See createFeatureVector for more details on structure
"""
# Store in feature class object
all_features_vector_dict = dict()
# For all motions
for motion_name in all_features_obj_dict:
feature_obj_list = all_features_obj_dict.get(motion_name)
all_adjective_labels_dict = dict()
feature_vector_list = list()
feature_label_list = list()
# For all objects
for bolt_feature_obj in feature_obj_list:
# Create feature vector
feature_vector, object_id = utilities.createFeatureVector(bolt_feature_obj, feature_name_list)
feature_vector_list.append(feature_vector)
feature_label_list.append(object_id)
# Create label dictionary
labels = bolt_feature_obj.labels
for adjective in labels:
# Check if it is the first time adjective added
if (all_adjective_labels_dict.has_key(adjective)):
adjective_array = all_adjective_labels_dict[adjective]
else:
adjective_array = list()
# Store array
adjective_array.append(labels[adjective])
all_adjective_labels_dict[adjective] = adjective_array
# Store all of the objects away
all_features_vector_dict[motion_name] = (np.array(feature_vector_list), feature_label_list)
return (all_features_vector_dict, all_adjective_labels_dict)
def bolt_obj_2_feature_vector(all_features_obj_dict, feature_name_list):
"""
Pull out PCA components from all data
For each object - pull out features and store in feature_obj
with the same structure as all_bolt_data
Dictionary - "tap", "slide", "slow_slide",
"thermal_hold", "squeeze"
Directly store the features into a vector
See createFeatureVector for more details on structure
"""
# DO PCA Calculations here
# Store in feature class object
all_features_vector_dict = dict()
# Store labels
for motion_name in all_features_obj_dict:
feature_obj_list = all_features_obj_dict.get(motion_name)
all_adjective_labels_dict = dict()
feature_vector_list = list()
# For all objects
for bolt_feature_obj in feature_obj_list:
# Create feature vector
feature_vector = utilities.createFeatureVector(bolt_feature_obj, feature_name_list)
feature_vector_list.append(feature_vector)
# Create label dictionary
labels = bolt_feature_obj.labels
for adjective in labels:
# Check if it is the first time adjective added
if (all_adjective_labels_dict.has_key(adjective)):
adjective_array = all_adjective_labels_dict[adjective]
else:
adjective_array = list()
# Store array
adjective_array.append(labels[adjective])
all_adjective_labels_dict[adjective] = adjective_array
# Store all of the objects away
all_features_vector_dict[motion_name] = np.array(feature_vector_list)
return (all_features_vector_dict, all_adjective_labels_dict)
"""
def true_false_results(predicted_labels, true_labels):
#FP = (predicted_labels - true_labels).tolist().count(1)
#FN = (predicted_labels - true_labels).tolist().count(-1)
#TP = (predicted_labels & true_labels).tolist().count(1)
#TN = ((predicted_labels | true_labels) ^ True).tolist().count(1)
return(TP, TN, FP, FN)
#def matthews_corr_coef(TP,TN,FP,FN):
try:
MCC = (TP*TN - FP*FN)/(np.sqrt(((TP+FP)*(TP+FN)*(TN+FP)*(TN+FN))))
except:
MCC = (TP*TN - FP*FN)/1
return (MCC)
"""
#def train_knn(train_vector, train_labels, test_vector, test_labels):
"""
train_knn - expects a vector of features and a nx1 set of
corresponding labels. Finally the number of
neighbors used for comparison
Returns a trained knn classifier
"""
"""