dis_r.append(dis)
dxy = []
for i, j in itertools.combinations(
range(0, training_data[m].shape[1], 2), 2):
dxy.append(training_data[m][r, i:i + 2] -
training_data[m][r, j:j + 2])
dxy_r.append(dxy)
dis_r = np.array(dis_r)
dxy_r = np.array(dxy_r)
dis_smth = []
dxy_eu = np.zeros([dataRange, dxy_r.shape[1]])
ang = np.zeros([dataRange - 1, dxy_r.shape[1]])
dxy_smth = []
ang_smth = []
for l in range(dis_r.shape[1]):
dis_smth.append(boxcar_center(dis_r[:, l], win_len))
for k in range(dxy_r.shape[1]):
for kk in range(dataRange):
dxy_eu[kk, k] = np.linalg.norm(dxy_r[kk, k, :])
if kk < dataRange - 1:
b_3d = np.hstack([dxy_r[kk + 1, k, :], 0])
a_3d = np.hstack([dxy_r[kk, k, :], 0])
c = np.cross(b_3d, a_3d)
ang[kk, k] = np.dot(
np.dot(np.sign(c[2]), 180) / np.pi,
math.atan2(
np.linalg.norm(c),
np.dot(dxy_r[kk, k, :], dxy_r[kk + 1, k, :])))
dxy_smth.append(boxcar_center(dxy_eu[:, k], win_len))
ang_smth.append(boxcar_center(ang[:, k], win_len))
dis_smth = np.array(dis_smth)
def bsoid_extract(data, fps):
"""
Extracts features based on (x,y) positions
:param data: list, csv data
:param fps: scalar, input for camera frame-rate
:return f_10fps: 2D array, extracted features
"""
win_len = np.int(np.round(0.05 / (1 / fps)) * 2 - 1)
feats = []
for m in range(len(data)):
logging.info(f'Extracting features from CSV file {m+1}...')
data_range = len(data[m])
dxy_r = []
dis_r = []
for r in range(data_range):
if r < data_range - 1:
dis = []
for c in range(0, data[m].shape[1], 2):
dis.append(
np.linalg.norm(data[m][r + 1, c:c + 2] -
data[m][r, c:c + 2]))
dis_r.append(dis)
dxy = []
for i, j in itertools.combinations(range(0, data[m].shape[1], 2),
2):
dxy.append(data[m][r, i:i + 2] - data[m][r, j:j + 2])
dxy_r.append(dxy)
dis_r = np.array(dis_r)
dxy_r = np.array(dxy_r)
dis_smth = []
dxy_eu = np.zeros([data_range, dxy_r.shape[1]])
ang = np.zeros([data_range - 1, dxy_r.shape[1]])
dxy_smth = []
ang_smth = []
for l in range(dis_r.shape[1]):
dis_smth.append(boxcar_center(dis_r[:, l], win_len))
for k in range(dxy_r.shape[1]):
for kk in range(data_range):
dxy_eu[kk, k] = np.linalg.norm(dxy_r[kk, k, :])
if kk < data_range - 1:
b_3d = np.hstack([dxy_r[kk + 1, k, :], 0])
a_3d = np.hstack([dxy_r[kk, k, :], 0])
c = np.cross(b_3d, a_3d)
ang[kk, k] = np.dot(
np.dot(np.sign(c[2]), 180) / np.pi,
math.atan2(
np.linalg.norm(c),
np.dot(dxy_r[kk, k, :], dxy_r[kk + 1, k, :])))
dxy_smth.append(boxcar_center(dxy_eu[:, k], win_len))
ang_smth.append(boxcar_center(ang[:, k], win_len))
dis_smth = np.array(dis_smth)
dxy_smth = np.array(dxy_smth)
ang_smth = np.array(ang_smth)
feats.append(np.vstack((dxy_smth[:, 1:], ang_smth, dis_smth)))
logging.info(
f'Done extracting features from a total of {len(data)} training CSV files.'
)
f_10fps = []
for n in range(0, len(feats)):
feats1 = np.zeros(len(data[n]))
for s in range(math.floor(fps / 10)):
for k in range(
round(fps / 10) + s, len(feats[n][0]), round(fps / 10)):
if k > round(fps / 10) + s:
feats1 = np.concatenate(
(feats1.reshape(feats1.shape[0], feats1.shape[1]),
np.hstack((
np.mean((feats[n][0:dxy_smth.shape[0],
range(k - round(fps / 10), k)]),
axis=1),
np.sum(
(feats[n][dxy_smth.shape[0]:feats[n].shape[0],
range(k - round(fps / 10), k)]),
axis=1))).reshape(len(feats[0]), 1)),
axis=1)
else:
feats1 = np.hstack(
(np.mean((feats[n][0:dxy_smth.shape[0],
range(k - round(fps / 10), k)]),
axis=1),
np.sum((feats[n][dxy_smth.shape[0]:feats[n].shape[0],
range(k - round(fps / 10), k)]),
axis=1))).reshape(len(feats[0]), 1)
logging.info(
'Done integrating features into 100ms bins from CSV file {n + 1}.'
)
f_10fps.append(feats1)
return f_10fps
def bsoid_feats(data: list, fps):
"""
Trains UMAP (unsupervised) given a set of features based on (x,y) positions
:param data: list of 3D array
:param fps: scalar, argument specifying camera frame-rate in LOCAL_CONFIG
:return f_10fps: 2D array, features
:return f_10fps_sc: 2D array, standardized/session features
"""
win_len = np.int(np.round(0.05 / (1 / fps)) * 2 - 1)
feats = []
for m in range(len(data)):
logging.info('Extracting features from CSV file {}...'.format(m + 1))
dataRange = len(data[m])
dxy_r = []
dis_r = []
for r in range(dataRange):
if r < dataRange - 1:
dis = []
for c in range(0, data[m].shape[1], 2):
dis.append(np.linalg.norm(data[m][r + 1, c:c + 2] - data[m][r, c:c + 2]))
dis_r.append(dis)
dxy = []
for i, j in itertools.combinations(range(0, data[m].shape[1], 2), 2):
dxy.append(data[m][r, i:i + 2] - data[m][r, j:j + 2])
dxy_r.append(dxy)
dis_r = np.array(dis_r)
dxy_r = np.array(dxy_r)
dis_smth = []
dxy_eu = np.zeros([dataRange, dxy_r.shape[1]])
ang = np.zeros([dataRange - 1, dxy_r.shape[1]])
dxy_smth = []
ang_smth = []
for l in range(dis_r.shape[1]):
dis_smth.append(boxcar_center(dis_r[:, l], win_len))
for k in range(dxy_r.shape[1]):
for kk in range(dataRange):
dxy_eu[kk, k] = np.linalg.norm(dxy_r[kk, k, :])
if kk < dataRange - 1:
b_3d = np.hstack([dxy_r[kk + 1, k, :], 0])
a_3d = np.hstack([dxy_r[kk, k, :], 0])
c = np.cross(b_3d, a_3d)
ang[kk, k] = np.dot(np.dot(np.sign(c[2]), 180) / np.pi,
math.atan2(np.linalg.norm(c),
np.dot(dxy_r[kk, k, :], dxy_r[kk + 1, k, :])))
dxy_smth.append(boxcar_center(dxy_eu[:, k], win_len))
ang_smth.append(boxcar_center(ang[:, k], win_len))
dis_smth = np.array(dis_smth)
dxy_smth = np.array(dxy_smth)
ang_smth = np.array(ang_smth)
feats.append(np.vstack((dxy_smth[:, 1:], ang_smth, dis_smth)))
logging.info('Done extracting features from a total of {} training CSV files.'.format(len(data)))
for n in range(0, len(feats)):
feats1 = np.zeros(len(data[n]))
for k in range(round(fps / 10), len(feats[n][0]), round(fps / 10)):
if k > round(fps / 10):
feats1 = np.concatenate((feats1.reshape(feats1.shape[0], feats1.shape[1]),
np.hstack((np.mean((feats[n][0:dxy_smth.shape[0],
range(k - round(fps / 10), k)]), axis=1),
np.sum((feats[n][dxy_smth.shape[0]:feats[n].shape[0],
range(k - round(fps / 10), k)]),
axis=1))).reshape(len(feats[0]), 1)), axis=1)
else:
feats1 = np.hstack((np.mean((feats[n][0:dxy_smth.shape[0], range(k - round(fps / 10), k)]), axis=1),
np.sum((feats[n][dxy_smth.shape[0]:feats[n].shape[0],
range(k - round(fps / 10), k)]), axis=1))).reshape(len(feats[0]), 1)
logging.info('Done integrating features into 100ms bins from CSV file {}.'.format(n + 1))
if n > 0:
f_10fps = np.concatenate((f_10fps, feats1), axis=1)
scaler = StandardScaler()
scaler.fit(feats1.T)
feats1_sc = scaler.transform(feats1.T).T
f_10fps_sc = np.concatenate((f_10fps_sc, feats1_sc), axis=1)
else:
f_10fps = feats1
scaler = StandardScaler()
scaler.fit(feats1.T)
feats1_sc = scaler.transform(feats1.T).T
f_10fps_sc = feats1_sc # scaling is important as I've seen wildly different stdev/feat between sessions
return f_10fps, f_10fps_sc