def main():
# basic setting
random.seed(opt['seed'])
np.random.seed(opt['seed'])
# create approach
isoh = IsoHAlgo(__DIM)
logger.info('creating isoh method done')
# load features
featurepath = os.path.join(opt['featurepath'], 'videos-features.h5')
features, mean_feature = load_features(featurepath)
# train isoh
train_features = load_features(featurepath, dtype='array')
isoh.learn_hash_function(train_features)
# generate binary codes
codes = isoh.generate_codes(features, mean_feature=mean_feature)
logger.info('generating codes done')
# load groundtruth and unlabeled-keys
gnds = load_groundtruth('test_groundtruth')
unlabeled_keys = get_video_id('unlabeled-data')
logger.info('loading gnds and unlabeled keys done. #query: {}'.format(len(gnds)))
# calculate map
map = calc_hamming_ranking(codes, unlabeled_keys, gnds)
logger.info('map: {:.4f}'.format(map))
logger.info('all done')
'''learning procedure'''
logger.info('start training procedure')
start_t = time.time()
db_image_codes, db_text_codes = dlfh_algo(train_labels=database_label)
'''out-of-sample extension'''
wx = linear_hash(database_image, db_image_codes)
wy = linear_hash(database_text, db_text_codes)
end_t = time.time() - start_t
'''start encoding'''
test_image_codes = np.sign(test_image.dot(wx))
test_text_codes = np.sign(test_text.dot(wy))
param = {'topk': [100]}
metrici2t = calc_hamming_ranking(test_image_codes, db_text_codes, test_label, database_label, param)
metrict2i = calc_hamming_ranking(test_text_codes, db_image_codes, test_label, database_label, param)
mapi2t = metrici2t['map']
mapt2i = metrict2i['map']
topprei2t = float(metrici2t['topkpre'][0])
toppret2i = float(metrict2i['topkpre'][0])
logger.info('map@i2t: {:>.4f}, map@t2i: {:>.4f}.'.format(mapi2t, mapt2i))
logger.info('precision@i2t: {:>.4f}, precision@t2i: {:>.4f}.'.format(topprei2t, toppret2i))