from data_pkg import data_fns as df
from functionals_pkg import feature_fns as ff
import numpy as np
import pickle
import os
print "TRAINING : "
path_videos = '/usr/local/data/sejacob/ANOMALY/data/UCSD/UCSD_Anomaly_Dataset.v1p2/UCSDped1/Train'
train_test = 'Train'
list_cuboids, all_cuboids = df.make_cuboids_of_videos(path_videos, train_test,
11, 11, 5)
mean = all_cuboids.mean(axis=0)
std = all_cuboids.std(axis=0)
all_cuboids = (all_cuboids - mean) / std
np.save(os.path.join('data_stored', 'cuboid_train_mean.npy'), mean)
np.save(os.path.join('data_stored', 'cuboid_train_std.npy'), std)
np.save(os.path.join('data_stored', 'all_cuboids_normed.npy'), all_cuboids)
with open(os.path.join('data_stored', 'list_cuboids.pkl'), 'wb') as f:
pickle.dump(list_cuboids, f)
len_local_feats = 13
feats_local = ff.make_all_local_feats(len_local_feats,
all_cuboids=list_cuboids,
mean=mean,
std=std)
print "LOADING CUBOIDS AND FEATS"
all_cuboids_normed = np.load(os.path.join('data_stored_temporal', 'all_cuboids_normed.npy'))
all_local_feats_normed = np.load(os.path.join('data_stored_temporal', 'all_local_feats_normed.npy'))
mean_data = np.load(os.path.join('data_stored_temporal', 'cuboid_train_mean.npy'))
std_data = np.load(os.path.join('data_stored_temporal', 'cuboid_train_std.npy'))
path_videos = '/usr/local/data/sejacob/ANOMALY/data/UCSD/UCSD_Anomaly_Dataset.v1p2/UCSDped1/Test'
train_test = 'Test'
size_axis = 12
n_frames = 5
list_cuboids_test, _, _ = df.make_cuboids_of_videos(path_videos, train_test, size_axis, size_axis, n_frames)
print "GET LOCAL FEATURE THRESHOLDS"
alpha_local, theta_local, mean_local, cov_inv_local = models.make_thresholds(all_local_feats_normed, 0.3)
del(all_local_feats_normed)
print "#######################"
print "LOCAL FEATURES"
print "#######################"
print "ALPHA_LOCAL:", alpha_local, " THETA_LOCAL:", theta_local
print "#######################"
thresholds_local = [alpha_local, theta_local]
print "SETTING LOCAL ANOMS"
list_cuboids_test = ff.set_anomaly_status_local_temporal(list_cuboids_test, thresholds_local, mean_local, cov_inv_local,
mean_data, std_data)
print "SAVING LOCAL ANOMS"
with open(os.path.join('data_stored_temporal', 'list_cuboids_test_local.pkl'), 'wb') as f:
print "$$$$$$$$$$$$$$$$$$$$$$$" print "FITTING THE KMEANS OBJECT" print "$$$$$$$$$$$$$$$$$$$$$$$" kmeans_obj = KMeans(n_clusters=int(metric['-n']), verbose=1,n_jobs=-1) kmeans_obj.fit(all_global_feats) print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" print "LOADING THE TRAIN LIST CUBOIDS:" print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" train = 'Train' list_cuboids_train, _, _ = df.make_cuboids_of_videos(path_videos_train, train, size_axis, size_axis, n_frames) print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" print "CREATE DICTIONARY:" print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" dictionary = ff.make_dictionary(list_cuboids_train, kmeans_obj, model,mean_data, std_data,mean_feats,std_feats) del(list_cuboids_train) print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" print "NUMBER OF DICTIONARY ENTRIES:", len(dictionary) print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$" print "Making rows into tuples" dictionary = [tuple(row) for row in dictionary]