diff --git a/S01cal.png b/S01cal.png
index 8cde599e24528f83cbc125db36a91a808b3b1ddc..2afbe6b9d355a0f194d880ee2c92a8f80a21dd86 100644
Binary files a/S01cal.png and b/S01cal.png differ
diff --git a/modules/datapipeline.py b/modules/datapipeline.py
index a05ec3487144a74ff4c451aeaf8053b76387a1a0..47d798546db06a2a4366e2ca23fc02a6dd87d010 100644
--- a/modules/datapipeline.py
+++ b/modules/datapipeline.py
@@ -2,7 +2,6 @@ import matplotlib.pyplot as plt
from datetime import datetime, timedelta
import pytz
import pandas as pd
-import dask.dataframe as dd
import tensorflow as tf
from os import environ, mkdir, makedirs, listdir, stat, walk
from sys import platform
diff --git a/regress_rr.py b/regress_rr.py
index c6c560e59f613a150fce774e09282662053135a4..c566abc3f88af87242a016e88b218d6ce3204574 100644
--- a/regress_rr.py
+++ b/regress_rr.py
@@ -1286,6 +1286,8 @@ def sens_rr_model(subject,
test_df_tmp = get_test_data(cal_df, activity_df, xsens_df, test_standing)
test_df = pd.concat([df for df in test_df_tmp['data']], axis=0)
+ ipdb.set_trace()
+
y_cols = ['sec', 'br', 'pss', 'cpm']
if use_tsfresh:
cal_df_list = []
diff --git a/single_subject_cal_plot.py b/single_subject_cal_plot.py
index c5017ab4f6f4c884bcca043db802fd9ac7ba96f2..c261bde1ae7f2197ef11b0c2e84ce18c93bcb01e 100644
--- a/single_subject_cal_plot.py
+++ b/single_subject_cal_plot.py
@@ -7,15 +7,22 @@ from os.path import join
from os import makedirs
import ipdb
import joblib
+from cycler import cycler
import pandas as pd
from config import *
-from regress_rr import get_activity_log
+from regress_rr import (
+ get_activity_log, load_and_sync_xsens, get_respiration_log,
+ get_cal_data
+)
+from regress_rr import get_test_data as get_raw_test_data
from modules.evaluations import Evaluation
from modules.digitalsignalprocessing import (
movingaverage, butter_lowpass_filter)
-from sklearn.preprocessing import PolynomialFeatures, LabelEncoder
+from sklearn.preprocessing import (
+ PolynomialFeatures, LabelEncoder, StandardScaler
+)
plt.close('all')
plt.rcParams.update({'figure.titlesize' : 8,
@@ -56,6 +63,172 @@ tsfresh_dir = join(
combi_strs = ['combi7.0-10.0-12.0-15.0-17.0']*3
+from matplotlib.animation import FuncAnimation
+cm = 1/2.54
+
+class AnimationPlotter():
+ def __init__(self, x, data:list, duration:int, buff_len:int,
+ vname='test.mp4'):
+ self.data = data
+ self.fig, self.axs = plt.subplots(3, 1, dpi=200,
+ figsize=(8*(4/3)*cm, 8*cm))
+ self.duration = duration
+ self.buff_len = buff_len
+ self.x0 = x[0]
+ self.x = x - x[0]
+ self.vname = vname
+
+ self.cm = [
+ cycler('color', sns.color_palette('dark')),
+ cycler('color', sns.color_palette('muted')),
+ cycler('color', sns.color_palette('pastel')),
+ ]
+
+ self.lines=[]
+ for i, ax in enumerate(self.axs):
+ ax.set_prop_cycle(self.cm[i])
+ ndim = self.data[i].shape[-1]
+ l = ax.plot(np.arange(0, buff_len), np.zeros((buff_len, ndim)))
+ self.lines.append(l)
+
+ ax.set_xlabel('Time (s)')
+ if i == 0:
+ # twin = ax.twinx()
+ ax.set_ylabel('Pressure')
+ # twin.set_ylabel('BR')
+ elif i == 1:
+ ax.set_ylabel('Accelerometer')
+ elif i == 2:
+ ax.set_ylabel('Gyroscope')
+ self.axs[1].legend(['x', 'y', 'z'], prop={'size': 6}, loc='upper left')
+ self.axs[0].set_xticklabels({})
+ self.axs[1].set_xticklabels({})
+
+ def animate(self, frame):
+ for i, m_line in enumerate(self.lines):
+ data = self.data[i]
+ sect = np.arange(frame, frame+self.buff_len)
+ if isinstance(m_line, list):
+ for j, m_l in enumerate(m_line):
+ m_l.set_data(self.x[sect], data[sect, j])
+ else:
+ m_line.set_data(self.x[sect], data[sect])
+ for ax in self.axs: ax.relim(); ax.autoscale_view(True, True, True);
+ plt.draw()
+ # return self.lines
+
+ def run(self):
+ ani = FuncAnimation(self.fig, self.animate, frames=self.duration,
+ repeat=False, interval=9)
+ ani.save(self.vname, writer='ffmpeg', codec='h264')
+ # plt.show()
+ return ani
+
+def do_animation(sect_duration, buffer_duration, sens_list=['imu', 'bvp']):
+ xsens_df = load_and_sync_xsens(sbj, sens_list=sens_list)
+ activity_df = get_activity_log(sbj).reset_index(drop=True)
+ event_df = get_respiration_log(sbj)
+
+ cal_df = get_cal_data(event_df, xsens_df)
+
+ # include standing or not
+ test_df_tmp = get_raw_test_data(cal_df, activity_df, xsens_df, 1)
+ test_df = pd.concat([df for df in test_df_tmp['data']], axis=0)
+
+ # cmap = sns.color_palette('Paired')
+
+ plot_strs = ['PSS']
+ acc_cols = ['acc_x', 'acc_y', 'acc_z']
+ gyro_cols = ['gyro_x', 'gyro_y', 'gyro_z']
+
+ def run_anim(df, vname):
+ df['PSS'] = StandardScaler().fit_transform(
+ df.PSS.values.reshape(-1, 1)).flatten()
+ df[acc_cols] = StandardScaler().fit_transform(df[acc_cols])
+ df[gyro_cols] = StandardScaler().fit_transform(df[gyro_cols])
+ x = df.sec.values
+
+ data = (df[plot_strs].values,
+ df[acc_cols].values,
+ df[gyro_cols].values,
+ )
+
+ anim = AnimationPlotter(x, data, duration, buff_len, vname=vname)
+ anim.run()
+
+ start_idx = 20*IMU_FS
+ buff_len = buffer_duration*IMU_FS
+ duration = sect_duration*IMU_FS
+
+ end_idx = start_idx + buff_len + duration
+
+ # get 3 calibration steps, 5, 10, 20
+ muted = sns.color_palette('muted')
+ pastel = sns.color_palette('pastel')
+ deep = sns.color_palette('deep')
+ m_figs, m_axs = [], []
+ # for i in [0, 2, 3, 4, 6]:
+ for i in [3]:
+ cpm = cal_df.cpm.iloc[i]
+ df = cal_df.data.iloc[i].iloc[start_idx:end_idx, :]
+ df['PSS'] = StandardScaler().fit_transform(
+ df.PSS.values.reshape(-1, 1)).flatten()
+ df[acc_cols] = StandardScaler().fit_transform(df[acc_cols])
+ df[gyro_cols] = StandardScaler().fit_transform(df[gyro_cols])
+ x = df.sec.values
+
+ data = (df[plot_strs].values,
+ df[acc_cols].values,
+ df[gyro_cols].values,
+ )
+
+ anim = AnimationPlotter(x, data, duration, buff_len,
+ vname='cal'+str(int(cpm))+'.mp4')
+ # anim.animate(1); plt.show()
+ # ipdb.set_trace()
+
+ '''
+ fig, ax = plt.subplots(3, 1)
+ ax[0].plot(df['PSS'], c='k', alpha=0.4)
+ twin = ax[0].twinx()
+ twin.plot(df['BR'], c='k')
+ ax[1].set_prop_cycle(cycler('color', deep))
+ ax[1].plot(StandardScaler().fit_transform(
+ df[['acc_x', 'acc_y', 'acc_z']]))
+
+ ax[2].set_prop_cycle(cycler('color', pastel))
+ ax[2].plot(
+ StandardScaler().fit_transform(df[['gyro_x', 'gyro_y', 'gyro_z']]),
+ )
+ m_figs.append(fig)
+ m_axs.append(ax)
+ '''
+ anim.run()
+
+ fmt = "%d/%m/%Y %H:%M:%S"
+ activity_sec = activity_df.Timestamps.map(
+ lambda x: datetime.strptime(x, fmt).timestamp()
+ )
+ seated_start, seated_end = activity_sec.iloc[2], activity_sec.iloc[3]
+ standing_start, standing_end = activity_sec.iloc[4], activity_sec.iloc[5]
+
+ # get seated no2, 7s
+ seated_df = test_df[
+ (test_df.sec > seated_start) & (test_df.sec < seated_end)
+ ]
+ seated_df = seated_df.iloc[start_idx:end_idx, :]
+ ipdb.set_trace()
+ run_anim(seated_df, 'seated.mp4')
+
+ # get standing no2, 7s
+ standing_df = test_df[
+ (test_df.sec > standing_start) & (test_df.sec < standing_end)
+ ]
+ standing_df = standing_df.iloc[start_idx:end_idx, :]
+ run_anim(standing_df, 'standing.mp4')
+
+ ipdb.set_trace()
+
def get_model(sens, sbj, cfg_id, combi):
m_dir = f'linreg_{sens}_rr_{sbj}_id{cfg_id}_{combi}'
mdl_fname = join(mdl_dir, sens+'_rr', str(cfg_id).zfill(2), m_dir, mdl_file)
@@ -77,6 +250,8 @@ if __name__ == '__main__':
standing_mask = activity_log['Activity'] == 'standing'
standing_log = activity_log[standing_mask]
+ do_animation(20, 15)
+
x_test, y_test, time = get_test_data()
y_out = {}
metrics_dict = {}
@@ -138,7 +313,7 @@ if __name__ == '__main__':
axs.set_title("Multi-Modal RR Predictions during Rest Standing and Sitting")
# axs[1].set_title("Absolute Error over Time")
- axs.set_xlabel("Time (indices)")
+ axs.set_xlabel("Time (sec)")
axs.set_ylabel("Respiration Rate (CPM)")
# axs[1].set_ylabel("Abs. Error")
fig.savefig('S01cal.png')