-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_dataset.py
267 lines (258 loc) · 10.5 KB
/
build_dataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
"""Build the features dataset from the raw antenna data."""
import argparse
import pathlib
import time
import numpy as np
import pandas as pd
import scipy.stats
import utils.load as load
import utils.features as features
def build_antenna_dataset(
data_path: pathlib.Path,
file_out: pathlib.Path,
nbins: int,
antenna_len_str: str,
) -> None:
# --------------
# Load data for each video into separate dataframes
(data_path / ".DS_Store").unlink(missing_ok=True)
file_out.parent.mkdir(parents=True, exist_ok=True)
data_paths = list(data_path.glob("*/CollectedData_Anna Hadjitofi.csv"))
dfs = [load.df_from_manual_labels(path) for path in data_paths]
# -----
# Compute features for each labelled bee in each frame (all variables in radians)
def loop_through_videos():
for j, df in enumerate(dfs):
print(j, "/", len(dfs))
individuals = set(i[0] for i in df.index)
for ind in individuals:
if ind == "dancer":
continue
df_ind = df.xs(ind)
for frame in df_ind:
bee_df_for_frame = df_ind[frame].astype(float)
dancer_df_for_frame = df.xs("dancer")[frame].astype(float)
# print("Bee id", ind, "Frame", frame)
angle_to_dancer = features.angle_of_nestmate_rel_to_dancer(
dancer_df_for_frame, bee_df_for_frame
) # Angle of nestmate relative to dancer
s_s_angle = features.s_s_angle(bee_df_for_frame) # Between scapes
l_angle, r_angle = features.angle_of_antennae_rel_to_body(
bee_df_for_frame, antenna_len_str
) # Angle of both antenna relative to body direction (straight ahead)
midpoint_angle = scipy.stats.circmean(
[l_angle, r_angle], high=np.pi, low=-np.pi
) # Antenna midpoint
recruit_gravity_angle = features.angle_of_bee_rel_to_up(
bee_df_for_frame
) # Angle of recruit relative to gravity
dancer_gravity_angle = features.angle_of_bee_rel_to_up(
dancer_df_for_frame
) # Angle of dancer relative to gravity
yield ind, frame, angle_to_dancer, s_s_angle, l_angle, r_angle, midpoint_angle, recruit_gravity_angle, dancer_gravity_angle
(
bee_ids,
frames,
angles_to_dancer,
s_s_angles,
l_antenna_angles,
r_antenna_angles,
midpoints_angles,
recruit_gravity_angles,
dancer_gravity_angles,
) = zip(*loop_through_videos())
angles_to_dancer_deg = np.rad2deg(angles_to_dancer)
remapped_angle_to_dancer_deg = features.angle_to_360(angles_to_dancer_deg)
# -----
# Store appropriate representations of the data
features_df = pd.DataFrame(
{
"bee_id": bee_ids,
"frame": frames,
"angle_to_dancer_rad": angles_to_dancer,
"angle_to_dancer_deg": angles_to_dancer_deg,
"remapped_angle_to_dancer_rad": np.deg2rad(remapped_angle_to_dancer_deg),
"remapped_angle_to_dancer_deg": remapped_angle_to_dancer_deg,
"s_s_rad": s_s_angles,
"s_s_deg": np.rad2deg(s_s_angles),
"l_antenna_rad": l_antenna_angles,
"l_antenna_deg": np.rad2deg(l_antenna_angles),
"r_antenna_rad": r_antenna_angles,
"r_antenna_deg": np.rad2deg(r_antenna_angles),
"antenna_midpoint_rad": midpoints_angles,
"antenna_midpoint_deg": np.rad2deg(midpoints_angles),
"recruit_angle_to_gravity_rad": recruit_gravity_angles,
"recruit_angle_to_gravity_deg": np.rad2deg(recruit_gravity_angles),
"dancers_angle_to_gravity_rad": dancer_gravity_angles,
"dancers_angle_to_gravity_deg": np.rad2deg(dancer_gravity_angles),
}
).dropna(
axis=0
) # Drops frames for individuals that dont have labels at that time
# -----
# Compute mean orientation of nestmate to their dancer at each time point
orientations_df = {}
longest_time_range, window_size = 0, 8
for name, group in features_df.groupby("bee_id"):
if group["angle_to_dancer_rad"].shape[0] < 10:
print(f"Skipping mean orientation for bee {name}. < 10 points")
continue
# For each group, compute the appropriate relative timestamps over the time
# that nestmate was recorded
time_range = features.get_time_scale(group["angle_to_dancer_rad"].values)
averaged_time = np.mean(
np.lib.stride_tricks.sliding_window_view(
time_range, window_shape=(window_size,)
),
axis=1,
).astype(int)
orientations_df[name] = {
"time": averaged_time,
"mean_orientation": features.compute_circmean_over_windows(
group["angle_to_dancer_rad"].values, window_size
),
"mean_orientation_gravity": features.compute_circmean_over_windows(
group["dancers_angle_to_gravity_rad"].values, window_size
),
}
longest_time_range = max(longest_time_range, max(time_range))
orientations_df = (
pd.DataFrame(orientations_df)
.T.explode(["time", "mean_orientation", "mean_orientation_gravity"])
.reset_index()
.rename(columns={"index": "bee_id"})
)
# -----
# Compute binned features
features_df["bin_idx_of_angle_to_dancer_deg"], bins = pd.cut(
features_df["angle_to_dancer_deg"],
bins=nbins,
retbins=True,
labels=range(nbins),
)
bins = np.roll(bins, shift=-1)
binned_df = (
features_df.groupby("bin_idx_of_angle_to_dancer_deg")
.apply(lambda group: features.compute_statistics_for_group(group))
.reset_index()
)
binned_df["bin_angle_to_dancer_deg"] = bins[
binned_df["bin_idx_of_angle_to_dancer_deg"]
]
binned_df["bin_angle_to_dancer_rad"] = np.deg2rad(
binned_df["bin_angle_to_dancer_deg"]
)
# Remap angle between -180 and 180 to 0 and 360
binned_df["remapped_bin_angle_to_dancer_deg"] = binned_df[
"bin_angle_to_dancer_deg"
].apply(lambda x: features.angle_to_360(x))
binned_df["remapped_bin_angle_to_dancer_rad"] = np.deg2rad(
binned_df["remapped_bin_angle_to_dancer_deg"]
)
# -----
# Compute gravity angles data (mean gravity angle of dancer for each nestmate)
dancers_gravity_df = {}
for name, group in features_df.groupby("bee_id"):
mean_dancer_angle = scipy.stats.circmean(
group["dancers_angle_to_gravity_rad"].values, high=np.pi, low=-np.pi
)
dancers_gravity_df[name] = mean_dancer_angle
dancers_gravity_df = (
pd.DataFrame(dancers_gravity_df, index=[0])
.T.reset_index()
.rename(columns={"index": "bee_id", 0: "mean_dancer_angle_to_gravity_rad"})
)
# -----
# Save data
if antenna_len_str == "mid_only":
file_out = str(file_out) + "-mid_antenna_len"
elif antenna_len_str == "full_only":
file_out = str(file_out) + "-full_antenna_len"
features_df.to_csv(f"{file_out}-features.csv", index=None)
orientations_df.to_csv(f"{file_out}-orientations.csv", index=None)
binned_df.to_csv(f"{file_out}-{nbins}-binned.csv", index=None)
dancers_gravity_df.to_csv(f"{file_out}-{nbins}-gravity.csv", index=None)
# --------------
def build_error_dataset(data_path: pathlib.Path, file_out: pathlib.Path, nbins: int):
"""Needs to be run after the above dataset has been built and we have simulated
the CX activity."""
# --------------
# Obtain list of np results files generated for each individual bee
files = list((data_path / "real_bee_dataset").glob("*.npz"))
(
errors_df,
reduced_errors_df,
binned_error_df,
consec_waggle_phases,
) = load.load_error_results_for_bees(files, nbins)
# -----
# Save data
experiment = data_path.name
errors_df.to_csv(f"{file_out}-{experiment}-{nbins}bins-errors.csv", index=None)
reduced_errors_df.to_csv(
f"{file_out}-{experiment}-{nbins}bins-reduced_errors.csv", index=None
)
binned_error_df.to_csv(
f"{file_out}-{experiment}-{nbins}bins-binned_errors.csv", index=None
)
consec_waggle_phases.to_csv(
f"{file_out}-{experiment}-{nbins}bins-consec_errors.csv", index=None
)
# --------------
if __name__ == "__main__":
# --------------
# User inputs
parser = argparse.ArgumentParser(description="Build antennal positioning dataset.")
parser.add_argument(
"--build_errors_ds",
action="store_true",
help="Build error dataset\
instead of antennae dataset.",
)
parser.add_argument(
"--data_path",
type=str,
default="Cropped-Anna Hadjitofi-2022-11-01",
help="Name of raw data folder. Defaults to 'Cropped-Anna Hadjitofi-2022-11-01'. \
For building errors dataset, this should be the path to the folder containing \
the experiments.",
)
parser.add_argument(
"--file_out",
type=str,
default=f"datasets/{time.strftime('%Y-%m-%d')}",
help="Resulting name (and path) to save dataset(s), without extension. \
Defaults to current date.",
)
parser.add_argument(
"--nbins",
type=int,
default=180,
help="Number of bins to use for binning angles to dancer. Defaults to 180.",
)
parser.add_argument(
"--antenna_len_str",
default="default",
type=str,
choices=["default", "mid_only", "full_only"],
help="Calculate antenna angle using mid length / bend or full length. The \
default uses base to tip if available and midpoint as fallback.",
)
args = vars(parser.parse_args())
# -----
args["file_out"] = pathlib.Path(args["file_out"])
# Build and save dataset
build_errors = args.pop("build_errors_ds")
if build_errors:
data_path = pathlib.Path(args["data_path"])
args.pop("antenna_len_str")
for experiment in data_path.glob("holoFly-*"):
print(experiment)
args["data_path"] = experiment
build_error_dataset(**args)
else:
args["data_path"] = pathlib.Path(
"dlc_antennae_tracking", args["data_path"], "labeled-data"
)
build_antenna_dataset(**args)
# --------------