def process(lst):
file_list = os.listdir(detail_path)
test_time = time.time()
startCount = 0
for file in file_list:
startCount += 1
if startCount < args.start or file == '_frame_num.txt' or file not in lst:
continue
file_to_read_fid = open(detail_path + file, 'r')
fid = file_to_read_fid.read()
fid = [int(i) for i in fid.split(' ')]
file_to_read_fid.close()
file_to_write_x = open(x_path + file, 'w')
file_to_write_y = open(y_path + file, 'w')
video_to_capture = file.split('.')[0] + '.avi'
vc = cv2.VideoCapture(video_path + video_to_capture) # 读入视频文件
if not vc.isOpened():
print('Open failure! exit')
exit(0)
rval = True
count = 0
index = 0
x_matrix = np.zeros((len(fid), joints_keys))
y_matrix = np.zeros((len(fid), joints_keys))
while rval and index < len(fid): # 循环读取视频帧
rval, frame = vc.read()
if count == fid[index]:
image = preprocess(frame, args.input_width, args.input_height)
joints = run(image)
for key in sorted(joints.keys()):
value = joints[key]
x_matrix[index][key] = value[0]
y_matrix[index][key] = value[1]
index += 1
count += 1
cv2.waitKey(1)
vc.release()
file_to_write_x.write(','.join(
str(i) for i in x_matrix.reshape(len(fid) * joints_keys).tolist()))
file_to_write_y.write(','.join(
str(i) for i in y_matrix.reshape(len(fid) * joints_keys).tolist()))
if (startCount % 100 == 0):
print(startCount, ": ", +time.time() - test_time)
file_to_write_x.close()
file_to_write_y.close()
print('total time: ', time.time() - test_time)
def get_prediction(img):
preprocessed = preprocess(img, input_width, input_height)
pafMat, heatMat = sess.run([
net.get_output(name=last_layer.format(stage=stage_level, aux=1)),
net.get_output(name=last_layer.format(stage=stage_level, aux=2))
],
feed_dict={'image:0': [preprocessed]})
heatMat, pafMat = heatMat[0], pafMat[0]
humans = estimate_pose(heatMat, pafMat)
return humans
def tf(text):
article = preprocess(text)
frequency, terms = calculateTermFrequency(article)
frequencySum = np.sum(frequency, axis=0)
# Calculate sentence weights
weightArray = []
for i in range(0, len(article)):
s = article[i]
weight = calculateSentenceWeight(s, frequencySum, terms)
weightArray.append((i, weight))
weightArray.sort(key=lambda a: a[1])
weightArray.reverse()
return weightArray
def getHumans(self, img):
preprocessed = preprocess(img, self.imgW, self.imgH)
run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
run_metadata = tf.RunMetadata()
pafMat, heatMat = self.sess.run([
self.net.get_output(
name=self.last_layer.format(stage=self.stage_level, aux=1)),
self.net.get_output(
name=self.last_layer.format(stage=self.stage_level, aux=2))
],
feed_dict={'image:0': [img]},
options=run_options,
run_metadata=run_metadata)
heatMat, pafMat = heatMat[0], pafMat[0]
humans = estimate_pose(heatMat, pafMat)
return humans
def get_eval_fn(model):
"""Return an evaluation function for server-side evaluation."""
# Load test data here to avoid the overhead of doing it in `evaluate` itself
_, test = tf.keras.datasets.mnist.load_data()
test_data, test_labels = test
# preprocessing
test_data, test_labels = common.preprocess(test_data, test_labels)
# The `evaluate` function will be called after every round
def evaluate(weights: fl.common.Weights):
model.set_weights(weights) # Update model with the latest parameters
loss, accuracy = model.evaluate(test_data, test_labels)
return loss, {"accuracy": accuracy}
return evaluate
def main():
data = np.loadtxt(args.input, delimiter=',')
X = common.preprocess(data[:, 1:-1])
y = data[:, -1].astype(np.uint8)
model = ng3.Ng3Model()
model.fit(X, y, monotonicity=[ng3.Ng3Model.NON_INCREASING, ng3.Ng3Model.NON_INCREASING])
df = model.predict(X)
for i, model_part in enumerate(model._models):
plt.figure()
plt.plot(model_part._xs, model_part._ys)
plt.title('Feature %d' % i)
plt.savefig('feature_%d.png' % i)
xx, yy = np.meshgrid(
np.linspace(min(X[:, 0]), max(X[:, 0]), 1000),
np.linspace(min(X[:, 1]), max(X[:, 1]), 1000))
Z = model.predict(np.hstack([xx.ravel().reshape(-1, 1), yy.ravel().reshape(-1, 1)]))
Z = Z.reshape(xx.shape)
plt.figure()
plt.contourf(xx, yy, Z)
plt.title('Decision surface')
plt.savefig('decision.png')
plt.scatter(X[y==0, 0], X[y==0, 1], c='b', marker='+')
plt.scatter(X[y==1, 0], X[y==1, 1], c='r', marker='+')
plt.savefig('decision_with_points.png')
thresh = np.percentile(df[y == 1], 100 * (1 - args.recall))
p, r = common.pr(df > thresh, y)
print('Threshold = %f' % thresh)
print(' p = %f' % p)
print(' r = %f' % r)
with open(args.output, 'wb') as output_file:
pickle.dump({
'model': model,
'thresh': thresh,
}, output_file)
def load_data_to_db(dir_prefix, db):
tweet_files = glob.glob(os.path.join(dir_prefix, '*.*'))
with open('../../anonymized_user_info_by_chunk_training.csv', 'r') as f:
reader = csv.reader(f)
next(reader, None)
users = {
row[0]: {
'age': row[1],
'num_tweets': row[2],
'gender': row[3],
'condition': row[4]
}
for row in reader
}
# user_records = [(un, float(users[un]['age']), int(users[un]['num_tweets']), users[un]['gender'], LABEL_IDS[users[un]['condition']]) for un in users.keys()]
#
# c = db.cursor()
# c.executemany('INSERT INTO users VALUES (?,?,?,?,?)', user_records)
# db.commit()
tweet_records = []
for file in tweet_files:
username = os.path.splitext(os.path.basename(file))[0]
print(username)
label = LABEL_IDS[users[username]['condition']]
tweet_file = open(file, 'r')
i = 0
for line in tweet_file:
tweet = json.loads(line)
i += 1
if valid_tweet(tweet):
t = preprocess(tweet['text'])
tweet_records.append((username, t))
tweet_file.close()
interpolation=cv2.INTER_LINEAR)
dx = (canvas.shape[1] - img_scaled.shape[1]) // 2
dy = (canvas.shape[0] - img_scaled.shape[0]) // 2
canvas[dy:dy + img_scaled.shape[0],
dx:dx + img_scaled.shape[1]] = img_scaled
img = canvas
elif args.zoom > 1.0:
img_scaled = cv2.resize(img,
None,
fx=args.zoom,
fy=args.zoom,
interpolation=cv2.INTER_LINEAR)
dx = (img_scaled.shape[1] - img.shape[1]) // 2
dy = (img_scaled.shape[0] - img.shape[0]) // 2
img = img_scaled[dy:img.shape[0], dx:img.shape[1]]
preprocessed = preprocess(img, args.input_width, args.input_height)
logging.debug('video process+')
pafMat, heatMat = sess.run([
net.get_output(
name=last_layer.format(stage=args.stage_level, aux=1)),
net.get_output(
name=last_layer.format(stage=args.stage_level, aux=2))
],
feed_dict={'image:0': [preprocessed]})
heatMat, pafMat = heatMat[0], pafMat[0]
logging.debug('video postprocess+')
t = time.time()
humans = estimate_pose(heatMat, pafMat)
images.extend(filenames)
break
if not path.exists(new_images_folder):
makedirs(new_images_folder)
counter = 0
img_data_arr = []
for filename in images:
rotation = generateRandomRotation()
img = Image.open(original_images_folder + '/' + filename)
#if (len(common.face_detect(img)) == 0):
img = img.rotate(rotation)
rotations_output_file.write(str(rotation) + '\n')
hog_1, color_histogram, img_face_rotation_data, guessed_rotation_by_faces = common.preprocess(
img)
img_data = np.concatenate(
(hog_1, hog_2, color_histogram, img_face_rotation_data,
[guessed_rotation_by_faces])).ravel()
img_data_arr.append(img_data)
img.save(new_images_folder + "/" + filename)
counter += 1
print filename
print img_face_rotation_data
print guessed_rotation_by_faces
print >> sys.stderr, str(counter)
#else:
# print filename + " has a face"
#if counter > 500:
# break
file_to_write = open(save_joints_path + save_file_name, 'w')
vc = cv2.VideoCapture(source_folder + file) # 读入视频文件
if not vc.isOpened():
print('Open failure! exit')
exit(0)
total = vc.get(cv2.CAP_PROP_FRAME_COUNT)
record_count = [0] * 19
record_sort = []
body = ""
rval = True
while rval: # 循环读取视频帧
rval, frame = vc.read()
joints = []
if frame is not None:
image = preprocess(frame, args.input_width,
args.input_height)
joints = run(image, show=False, trg_len=15)
# print(len(joints), joints)
for key in sorted(joints.keys()):
value = joints[key]
body += (str(key) + ':')
body += (str(value[0]) + ',' + str(value[1]) + ' ')
body += '\n'
record_count[len(joints)] += 1
record_sort.append(len(joints))
cv2.waitKey(1)
vc.release()
title = ' '.join(str(i) for i in record_count) + '\n'
title += (' '.join(str(i) for i in record_sort) + '\n')
file_to_write.write(title)
file_to_write.write(body)
heatmaps_tensor = tf.get_default_graph().get_tensor_by_name('Mconv7_stage6_L2/BiasAdd:0')
pafs_tensor = tf.get_default_graph().get_tensor_by_name('Mconv7_stage6_L1/BiasAdd:0')
t3 = time.time()
print("Time {} for tensor creation".format(t3 - t2))
with tf.Session() as sess:
while cam.isOpened():
success, image = cam.read()
t4 = time.time()
print("Time {} for reading cam".format(t4 - t3))
if not success:
continue
image = preprocess(image, 656, 368)
t5 = time.time()
print("Time {} for preprocessed image".format(t5 - t4))
# The session runner is really slow
heatMat, pafMat = sess.run([heatmaps_tensor, pafs_tensor], feed_dict={
inputs: image
})
t6 = time.time()
print("Time {} for session runner".format(t6 - t5))
heatMat, pafMat = heatMat[0], pafMat[0]
humans = estimate_pose(heatMat, pafMat)
images.extend(filenames)
break
if not path.exists(new_images_folder):
makedirs(new_images_folder)
counter = 0
img_data_arr = []
for filename in images:
rotation = generateRandomRotation()
img = Image.open(original_images_folder + '/' + filename)
#if (len(common.face_detect(img)) == 0):
img = img.rotate(rotation)
rotations_output_file.write(str(rotation) + '\n')
hog_1, color_histogram, img_face_rotation_data, guessed_rotation_by_faces = common.preprocess(img)
img_data = np.concatenate((hog_1, hog_2, color_histogram, img_face_rotation_data, [guessed_rotation_by_faces])).ravel()
img_data_arr.append(img_data)
img.save(new_images_folder + "/" + filename)
counter += 1
print filename
print img_face_rotation_data
print guessed_rotation_by_faces
print >> sys.stderr, str(counter)
#else:
# print filename + " has a face"
#if counter > 500:
# break
def evaluate(X, model, min_support=DEFAULT_MIN_SUPPORT):
y_hat = model['model'].predict(common.preprocess(X)) > model['thresh']
return common.convolve_left(y_hat,
np.ones(10).astype(np.uint8)) > min_support
