def save_specific(to_save, full=False, gray=False):
prep = Preprocessor(crop_style=1, gray=gray, should_normalize=False)
for i in to_save:
if full:
save_full(i, trial[i])
else:
ball, car = prep.process_frame(trial[i])
save(i, ball, car, gray)
def generateSummary(PATH, SIZE):
p = Preprocessor(PATH, 1).parse()
scorer = SentenceScoreCalculator(p)
summary = sorted(
sorted(range(p.sCount), key=lambda s: scorer.rank(s),
reverse=True)[0:SIZE])
return map(lambda s: p.sentences[s], summary)
def init():
prep = Preprocessor()
trial = np.load('data/trial_{}.npy'.format(1))
ball_crop = prep.get_ball_batch(trial).astype(np.float32)
global ball_mean, ball_mse_mean, ball_mse_std
ball_mean = np.mean(ball_crop, axis=0)
mses = np.mean(np.square(ball_crop - ball_mean), axis=(1, 2, 3))
ball_mse_mean = mses.mean()
ball_mse_std = mses.std()
def main():
prep = Preprocessor(crop_style=1, should_normalize=False)
for i, frame in enumerate(trial):
print(i)
ball, car = prep.get_ball_and_car(frame)
show(car)
cv2.waitKey(0)
keys = key_check()
while not (len(keys) == 1 and ('0' in keys or '1' in keys)):
time.sleep(0.1)
keys = key_check()
if '1' in keys:
save(i, ball, car)
cv2.destroyAllWindows()
def __init__(
self,
crop_style=0,
gray=False,
frame_time=1 / 8,
read_only=False,
normalize=True,
buffer_size=3,
field='champions',
reward_func=lambda x: 1,
):
self.field = field
self.crop_style = crop_style
if crop_style is None:
self.prep = None
else:
self.prep = Preprocessor(crop_style,
gray=gray,
should_normalize=normalize,
field=field)
if gray:
self.ball_obs_dims = self.prep.ball_dims + (buffer_size, )
self.car_obs_dims = self.prep.car_dims + (buffer_size, )
else:
self.ball_obs_dims = self.prep.ball_dims + (3 * buffer_size, )
self.car_obs_dims = self.prep.car_dims + (3 * buffer_size, )
self.bboxes = (get_screen_bbox(), )
self.next_frame = time.time()
self.read_only = read_only
self.frame_time = frame_time
self.get_reward = reward_func
self.gray = gray
self.frame_buffer = deque(maxlen=buffer_size) # Save last three frames
self.frame_times = []
self.obs_times = []
def generateSummary(PATH, NGRAMS, CLUSTERS, TYPE):
# Preprocessing and Term Selection
p = Preprocessor(PATH, NGRAMS).parse()
counter = WordCounter(p.sCount).count(p.processed)
# Term weighting and feature selection
features = buildFeatures(p, counter, TYPE)
# Sentence clustering
kmeans = KMeans(
n_clusters=CLUSTERS,
#n_jobs=4,
precompute_distances=True,
init="k-means++",
random_state=500,
tol=1e-10,
copy_x=True)
kmeans.fit(features)
#Sentence selection
summary = sentenceSelect(kmeans, p, features)
return map(lambda s: p.sentences[s], sorted(summary))
help="Enables the stopwords' filter (using nltk's stopwords list).",
action="store_true")
parser.add_argument("-t",
help="Enables the stopwords' filter (using pos tag).",
action="store_true")
parser.add_argument("-u", help="Enables the url filter.", action="store_true")
parser.add_argument("-n", help="Enables the not adapter.", action="store_true")
args = parser.parse_args()
######### Object creation
# timer used for timing
timer = Timer()
# a preprocessor
preprocess = Preprocessor()
print ""
print ""
# Classifier selection
if (args.classifier == 'svm'):
from classifier.svmClassifier import SVMClassifier
print "Classifier: SVM Classifier."
classifier = SVMClassifier()
else:
if (args.classifier == 'bayes'):
from classifier.bayesianClassifier import BayesianClassifier
print "Classifier: Bayesian Classifier."
classifier = BayesianClassifier()
else:
from classifier.shortTextClassifier import ShortTextClassifier
# print(i, mses[index])
# show(trial[index])
# # print(i, mses[i])
# # show(trial[i])
def show_gray(img):
cv2.imshow('cropped', img)
cv2.waitKey(0)
cv2.destroyAllWindows()
gray_minframe = gray_mses.argmin()
gray_maxframe = gray_mses.argmax()
show_gray(gray_crop[gray_minframe].reshape(gray_mean.shape))
show_gray(gray_mean)
show_gray(gray_crop[gray_maxframe].reshape(gray_mean.shape))
show_gray(gray_mean)
if __name__ == '__main__':
prep = Preprocessor(crop_style=1)
# 13 trials
for trial_num in range(7):
print('Displaying trial {}'.format(trial_num))
trial = np.load('data/champions/trial_{}.npy'.format(trial_num))
show(trial[0])
ball, car = prep.get_ball_and_car(trial[0])
show(ball)
show(car)
# ball_crop(trial)