def tracksSpeedsPerSpike(exp_files):
sns.set()
sns.set_context('paper')
for i, exp_tuple in enumerate(exp_files):
dir_name = path.dirname(exp_tuple[0])
cur_artifacts = Artifacts(expLocation=dir_name)
tracks_by_pulse = cur_artifacts.getArtifact('tracks_by_pulse')
for key in tracks_by_pulse.keys():
pulse_df = pd.DataFrame()
current_tracks = tracks_by_pulse[key]
spikes_speeds = np.array([t._tracksSpeeds for t in current_tracks])
# Getting rid start/end time artifacts.
spikes_speeds = spikes_speeds[:, 2:-2]
df = pd.DataFrame({
'Time':
tuple(range(0, spikes_speeds.shape[1])) * 0.5 *
spikes_speeds.shape[0],
'Speed':
np.reshape(spikes_speeds, np.size(spikes_speeds))
})
fig = sns.lineplot(x='time', y='speed', data=df, ci=None)
fig.set_axes_label('Speed [au]', 'Time [s]')
pass
def main():
folder = '/home/itskov/Temp/Chris'
#folder = '/mnt/storageNASRe/ChristianData/ChrisNewTracks'
conds = ['LTAP.', 'LTAV.', 'STAP.', 'STAV.', 'NAIVE']
df = None
for cond in conds:
print('Cond: ' + cond)
for i, filename in enumerate(Path(folder).rglob(cond + "*")):
print('Working on: %s' % (filename, ))
currentArt = Artifacts(expLocation=filename)
current_roi = currentArt.getArtifact('roi')
if current_roi is None:
continue
if current_roi['wormCount'] < 20:
print('Not enough worms: %d. Continuing.' %
(current_roi['wormCount'], ))
continue
TIME = 2160
frames = range(1, len(current_roi['arrivedFrac']) + 1)
arrived_frac = np.maximum(0, current_roi['arrivedFrac'][0:TIME])
current_df = pd.DataFrame({
'frame': range(1, TIME + 1),
'arrived_frac': arrived_frac,
'cond': cond
})
df = current_df if df is None else pd.concat(
(df, current_df), ignore_index=True)
sns.set()
#sns.lineplot(x='frame',y='arrived_frac', hue='cond', data=df, estimator=np.median, ci=68, n_boot=1000)
df.to_csv('/home/itskov/workspace/fraction_in_time.csv')
#plt.ylim([-0.001, 0.15])
#plt.xlabel('Frame [2hz]')
#plt.ylabel('Arrived Fraction')
#plt.show()
pass
def tracksSpeedsForAllSpikes(exp_files, paper=False, show=True):
#sns.set()
if not paper:
sns.set_context('talk')
plt.style.use('dark_background')
else:
sns.set_context('paper')
df = pd.DataFrame({'Frame': [], 'Speed': [], 'Activation Duration': []})
for i, exp_tuple in enumerate(exp_files):
dir_name = path.dirname(exp_tuple[0])
cur_artifacts = Artifacts(expLocation=dir_name)
tracks_for_all_pulses = cur_artifacts.getArtifact(
'tracks_for_all_spikes')
spikes_speeds = np.array(
[t._tracksSpeeds for t in tracks_for_all_pulses])
# Getting rid start/end time artifacts.
spikes_speeds = spikes_speeds[:, 2:-2]
norm_factor = np.reshape(np.mean(spikes_speeds[:, 0:10], axis=1),
(spikes_speeds.shape[0], 1))
spikes_speeds = spikes_speeds - norm_factor
current_df = pd.DataFrame({
'Time':
tuple(np.array(range(0, spikes_speeds.shape[1])) * 0.5) *
spikes_speeds.shape[0],
'Speed':
np.reshape(spikes_speeds, np.size(spikes_speeds)),
'Activation Duration':
exp_tuple[1]
})
cp = sns.cubehelix_palette(8)
ax = sns.lineplot(x='Time',
y='Speed',
data=current_df,
ci=95,
color=cp[5],
linewidth=2)
plt.gca().grid(alpha=0.2)
ax.set(xlabel='Time [s]', ylabel='Normalized Speed [au]')
plt.gca().grid(alpha=0.2)
df = pd.concat((df, current_df))
pass
ax = sns.lineplot(x='Time',
y='Speed',
hue='Activation Duration',
data=df,
ci=None,
linewidth=2,
alpha=0.75)
plt.gca().grid(alpha=0.2)
ax.set(xlabel='Time [s]', ylabel='Speed [au]')
plt.gca().grid(alpha=0.2)
if show:
plt.show()
pass
def day_summary_plots(exp_pairs,
titles,
legends,
paper=False,
show=True,
output_file=None):
number_of_exps = len(exp_pairs)
fig, axs = plt.subplots(2,
int(np.ceil(number_of_exps / 2)),
figsize=(16, 4))
plt.subplots_adjust(wspace=0.3, top=0.93, hspace=0.51)
if not paper:
plt.style.use("dark_background")
sns.set_context("talk")
else:
sns.set_context("paper")
axs = np.atleast_2d(axs)
df = pd.DataFrame({'Time': [], 'ATR+': [], 'ATR-': [], 'Type': []})
for i, exp_pair in enumerate(exp_pairs):
# ROI plot
plt.sca(axs[int(i >= 3)][int(i % 3)])
first_exp_art = Artifacts(expLocation=exp_pair[0])
second_exp_art = Artifacts(expLocation=exp_pair[1])
PairWiseRoi_(['ATR+', 'ATR- (Ctrl)'], [
first_exp_art.getArtifact('roi'),
second_exp_art.getArtifact('roi')
],
showShow=False,
paper=paper,
show_count=False,
freq=120)
#plt.title(titles[i])
#plt.legend(fontsize='small', title_fontsize='40')
first_mean_speed = np.mean(first_exp_art.getArtifact('speed')['speed'])
second_mean_speed = np.mean(
second_exp_art.getArtifact('speed')['speed'])
first_mean_proj = np.mean(first_exp_art.getArtifact('proj')['proj'])
second_mean_proj = np.mean(second_exp_art.getArtifact('proj')['proj'])
first_max_roi = np.mean(
first_exp_art.getArtifact('roi')['arrivedFrac'])
second_max_roi = np.mean(
second_exp_art.getArtifact('roi')['arrivedFrac'])
cur_df = pd.DataFrame({
'Time':
i,
'ATR+': [first_mean_speed, first_mean_proj, first_max_roi],
'ATR-': [second_mean_speed, second_mean_proj, second_max_roi],
'Type': ['Speed', 'Projection', 'Roi']
})
df = pd.concat((df, cur_df), ignore_index=True)
#fig.tight_layout()
#fig.savefig('/home/itskov/Dropbox/dayfigs.png')
if show:
plt.show()
return df
from Behavior.Tools.Artifacts import Artifacts
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
art_120a = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_ATR_TRAIN_65M_D120_NO_IAA3x5.avi_12.57.03/')
art_120na = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_NO_ATR_TRAIN_65M_D120_NO_IAA3x5.avi_12.56.12/')
art_180a = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_ATR_TRAIN_65M_D180_NO_IAA3x5.avi_15.29.04/')
art_180na = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_NO_ATR_TRAIN_65M_D180_NO_IAA3x5.avi_15.28.03/')
art_240a = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_ATR_TRAIN_65M_D240_NO_IAA3x5.avi_17.28.52/')
art_240na = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_NO_ATR_TRAIN_65M_D240_NO_IAA3x5.avi_17.28.04/')
art_300a = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_ATR_TRAIN_65M_D300_NO_IAA3x5.avi_20.12.12/')
art_300na = Artifacts(expLocation='/mnt/storageNASRe/tph1/Results/22-Jan-2020/TPH_1_NO_ATR_TRAIN_65M_D300_NO_IAA3x5.avi_20.11.34/')
art_120a_frac = art_120a.getArtifact('roi')['arrivedFrac'][4499]
art_120na_frac = art_120na.getArtifact('roi')['arrivedFrac'][4499]
art_180a_frac = art_180a.getArtifact('roi')['arrivedFrac'][4499]
art_180na_frac = art_180na.getArtifact('roi')['arrivedFrac'][4499]
art_240a_frac = art_240a.getArtifact('roi')['arrivedFrac'][4499]
art_240na_frac = art_240na.getArtifact('roi')['arrivedFrac'][4499]
art_300a_frac = art_300a.getArtifact('roi')['arrivedFrac'][4499]
art_300na_frac = art_300na.getArtifact('roi')['arrivedFrac'][4499]
def scatterEnhacnment(pairsDf, paper=False, figNames=None):
MAX_TIME = 3750
plotDfArrivedFrac = pd.DataFrame({'Strain': [], 'ATR+': [], 'ATR-': []})
plotDfProjection = pd.DataFrame({'Strain': [], 'ATR+': [], 'ATR-': []})
plotDfSpeed = pd.DataFrame({'Strain': [], 'ATR+': [], 'ATR-': []})
linePlotDf = pd.DataFrame({
'Strain': [],
'time': [],
'FrationArrived': [],
'Exp': []
})
print(pairsDf.shape)
for i in range(pairsDf.shape[0]):
try:
current_row = pairsDf.iloc[i]
files = np.array(current_row['files'])
noAtrInd = np.array([file.find('NO_ATR') >= 0 for file in files])
if np.sum(noAtrInd) != 1:
print('*** Error: Invalid pair ***')
else:
atrFile = files[np.logical_not(noAtrInd)][0]
noAtrFile = files[noAtrInd][0]
print(atrFile)
print(noAtrFile)
atrRoi = Artifacts(expLocation=atrFile).getArtifact('roi')
noAtrRoi = Artifacts(expLocation=noAtrFile).getArtifact('roi')
if atrRoi['wormCount'] < 30 or noAtrRoi['wormCount'] < 30:
print('Not enough worms. Continuing.')
continue
plotDfArrivedFrac = plotDfArrivedFrac.append(
{
'Strain': current_row['Strain'],
'ATR+': np.minimum(atrRoi['arrivedFrac'][MAX_TIME], 1),
'ATR-': np.minimum(noAtrRoi['arrivedFrac'][MAX_TIME],
1),
'ExpId': i
},
ignore_index=True)
atrProj = Artifacts(
expLocation=atrFile).getArtifact('proj')['proj']
noAtrProj = Artifacts(
expLocation=noAtrFile).getArtifact('proj')['proj']
plotDfProjection = plotDfProjection.append(
{
'Strain': current_row['Strain'],
'ATR+': np.mean(atrProj),
'ATR-': np.mean(noAtrProj),
'ExpId': i
},
ignore_index=True)
atrSpeed = Artifacts(
expLocation=atrFile).getArtifact('speed')['speed']
noAtrSpeed = Artifacts(
expLocation=noAtrFile).getArtifact('speed')['speed']
plotDfSpeed = plotDfSpeed.append(
{
'Strain': current_row['Strain'],
'ATR+': np.mean(atrSpeed),
'ATR-': np.mean(noAtrSpeed),
'ExpId': i
},
ignore_index=True)
print(
"Arrival vals: ATR+:%f, ATR-:%f, Projection mean: ATR+:%f, ATR-:%f, Speed mean ATR+:%f, ATR-:%f"
% (atrRoi['arrivedFrac'][MAX_TIME],
noAtrRoi['arrivedFrac'][MAX_TIME], np.mean(atrProj),
np.mean(noAtrProj), np.mean(atrSpeed),
np.mean(noAtrSpeed)))
print(atrRoi['arrivedFrac'][0:MAX_TIME].shape)
print(np.array(list(range(MAX_TIME))).shape)
currentLinePlot = pd.DataFrame({
'Strain':
current_row['Strain'],
'time':
np.array(list(range(MAX_TIME))) * 0.5,
'FractionArrived':
atrRoi['arrivedFrac'][0:MAX_TIME],
'Exp':
'ATR'
})
linePlotDf = pd.concat((linePlotDf, currentLinePlot))
currentLinePlot = pd.DataFrame({
'Strain':
current_row['Strain'],
'time':
np.array(list(range(MAX_TIME))) * 0.5,
'FractionArrived':
noAtrRoi['arrivedFrac'][0:MAX_TIME],
'Exp':
'NO ATR'
})
linePlotDf = pd.concat((linePlotDf, currentLinePlot))
except Exception as e:
raise e
#print(plotDf)
#DEBUG
#plotDfArrivedFrac.to_pickle('/home/itskov/Dropbox/tempBundle_arrival.pkl')
#plotDfProjection.to_pickle('/home/itskov/Dropbox/tempBundle_proj.pkl')
#plotDfSpeed.to_pickle('/home/itskov/Dropbox/tempBundle_speed.pkl')
#linePlotDf.to_pickle('/home/itskov/Dropbox/tempBundle2.pkl')
#DEBUG
if paper == False:
plt.style.use("dark_background")
sns.set_context('talk')
cp = sns.dark_palette("purple", 7)
else:
cp = sns.cubehelix_palette(8, start=.5, rot=-.75)
if paper == False:
ax = sns.scatterplot(x='ATR-',
y='ATR+',
data=plotDfArrivedFrac,
linewidth=0,
alpha=0.85,
color=cp[6])
ax.plot([0, 1.1], [0, 1.1], ":")
else:
ax = sns.scatterplot(x='ATR-',
y='ATR+',
data=plotDfArrivedFrac,
linewidth=0,
alpha=0.85,
color=cp[6])
ax.plot([0, 1.1], [0, 1.1], ":", color='k')
plt.xlim(0, 1.1)
plt.ylim(0, 1.1)
plt.gca().grid(alpha=0.2)
plt.title('Arrival Fracion, n = %d' % (plotDfArrivedFrac.shape[0], ),
loc='left')
if figNames is None:
plt.show()
else:
plt.gcf().savefig(figNames[0], format='svg')
# Plotting the projection plot.
#plt.style.use("dark_background")
#sns.set_context('talk')
#cp = sns.dark_palette("purple", 7)
ax = sns.scatterplot(x='ATR-',
y='ATR+',
data=plotDfProjection,
linewidth=0,
alpha=0.85,
color=cp[6])
xmin = np.min(plotDfProjection['ATR-'])
xmax = np.max(plotDfProjection['ATR-'])
ymin = np.min(plotDfProjection['ATR-'])
ymax = np.max(plotDfProjection['ATR+'])
if paper == False:
ax.plot([0, 1], [0, 1], ":")
else:
ax.plot([0, 1], [0, 1], ":", color='k')
plt.xlim(0, 0.35)
plt.ylim(0, 0.35)
plt.gca().grid(alpha=0.2)
plt.locator_params(axis='y', nbins=6)
plt.locator_params(axis='x', nbins=6)
plt.title('Mean Projection, n = %d' % (plotDfArrivedFrac.shape[0], ),
loc='left')
if figNames is None:
plt.show()
else:
plt.gcf().savefig(figNames[1], format='svg')
# Plotting the speed plot.
#plt.style.use("dark_background")
#sns.set_context('talk')
#cp = sns.dark_palette("purple", 7)
ax = sns.scatterplot(x='ATR-',
y='ATR+',
data=plotDfSpeed,
linewidth=0,
alpha=0.85,
color=cp[6])
xmin = np.min(plotDfSpeed['ATR-'])
xmax = np.max(plotDfSpeed['ATR-'])
ymin = np.min(plotDfSpeed['ATR+'])
ymax = np.max(plotDfSpeed['ATR+'])
if paper == False:
ax.plot([0, 0.0018], [0, 0.0018], ":")
else:
ax.plot([0, 0.0018], [0, 0.0018], ":", color='k')
plt.xlim(0.0008, 0.0018)
plt.ylim(0.0008, 0.0018)
plt.gca().grid(alpha=0.2)
plt.locator_params(axis='y', nbins=6)
plt.locator_params(axis='x', nbins=6)
plt.title('Mean Speed, n = %d' % (plotDfSpeed.shape[0], ), loc='left')
if figNames is None:
plt.show()
else:
plt.gcf().savefig(figNames[2], format='svg')
#ax = sns.lineplot(x='time', y='FractionArrived', hue='Exp', data=linePlotDf, ci=68, estimator=np.median)
#plt.gca().grid(alpha=0.2)
#ax.set(xlabel='Time [s]')
#plt.show()
return (plotDfArrivedFrac, plotDfProjection, plotDfSpeed)
def day_summary_plots(exp_pairs,
titles,
legends,
paper=False,
show=True,
output_file=None):
number_of_exps = len(exp_pairs)
fig, axs = plt.subplots(number_of_exps, 3, figsize=(10, 14))
plt.subplots_adjust(wspace=0.36,
left=0.09,
bottom=0.11,
right=0.98,
top=0.96,
hspace=0.36)
if not paper:
plt.style.use("dark_background")
sns.set_context("talk")
else:
sns.set_context("paper")
axs = np.atleast_2d(axs)
df = pd.DataFrame({'Value': [], 'Type': [], 'Cond': [], 'Exp': []})
for i, exp_pair in enumerate(exp_pairs):
first_exp_art = Artifacts(expLocation=exp_pair[0])
second_exp_art = Artifacts(expLocation=exp_pair[1])
plt.sca(axs[int(i)][0])
PairWiseRoi_(['ATR+', 'ATR- (Ctrl)'], [
first_exp_art.getArtifact('roi'),
second_exp_art.getArtifact('roi')
],
showShow=False,
paper=paper,
show_count=False,
freq=120)
df = df.append(
{
'Value': np.max(
first_exp_art.getArtifact('roi')['arrivedFrac']),
'Type': 'Max Roi',
'Cond': 'ATR+',
'Exp': i
},
ignore_index=True)
df = df.append(
{
'Value': np.max(
second_exp_art.getArtifact('roi')['arrivedFrac']),
'Type': 'Max Roi',
'Cond': 'ATR-',
'Exp': i
},
ignore_index=True)
plt.sca(axs[int(i)][1])
first_speed = (first_exp_art.getArtifact('speed')['speed'])
second_speed = (second_exp_art.getArtifact('speed')['speed'])
sns.kdeplot(first_speed, shade=True, label='ATR+')
ax = sns.kdeplot(second_speed, shade=True, label='ATR- (Control)')
plt.gca().grid(alpha=0.2)
ax.set(xlabel="Speed [au / sec]", ylabel="Density")
plt.locator_params(nbins=5)
df_cur = pd.DataFrame(
{
'Value': np.mean(first_speed),
'Type': 'Speed [au]',
'Cond': 'ATR+',
'Exp': i
},
index=[0])
df = pd.concat((df, df_cur), ignore_index=True)
df_cur = pd.DataFrame(
{
'Value': np.mean(second_speed),
'Type': 'Speed [au]',
'Cond': 'ATR-',
'Exp': i
},
index=[0])
df = pd.concat((df, df_cur), ignore_index=True)
plt.sca(axs[int(i)][2])
first_proj = (first_exp_art.getArtifact('proj')['proj'])
second_proj = (second_exp_art.getArtifact('proj')['proj'])
sns.kdeplot(first_proj, shade=True, label='ATR+')
ax = sns.kdeplot(second_proj, shade=True, label='ATR- (Control)')
plt.gca().grid(alpha=0.2)
ax.set(xlabel="Projection", ylabel="Density")
plt.locator_params(nbins=5)
df_cur = pd.DataFrame(
{
'Value': np.mean(first_proj),
'Type': 'Projection',
'Cond': 'ATR+',
'Exp': i
},
index=[0])
df = pd.concat((df, df_cur), ignore_index=True)
df_cur = pd.DataFrame(
{
'Value': np.mean(second_proj),
'Type': 'Projection',
'Cond': 'ATR-',
'Exp': i
},
index=[0])
df = pd.concat((df, df_cur), ignore_index=True)
if show:
plt.show()
return df
def meanProjectionMeanWithIaa(firstGroupName, secondGroupName, rootPath):
firstExperimentsFiles = list(
Path(rootPath).rglob(firstGroupName + '*/exp.npy'))
secondExperimentsFiles = list(
Path(rootPath).rglob(secondGroupName + '*/exp.npy'))
interestingTimePoints = [0, 1000, 2000, 3000, 4000]
firstGroupResults = np.zeros(
(len(firstExperimentsFiles), len(interestingTimePoints)))
secoundGroupResults = np.zeros(
(len(secondExperimentsFiles), len(interestingTimePoints)))
print('First Group:')
for i, filename in enumerate(firstExperimentsFiles):
if 'Nov' not in str(filename):
continue
if '03-Nov' in str(filename):
continue
# Getting the dirname of the experiment.
expDirName = path.dirname(filename)
arts = Artifacts(expLocation=expDirName)
roiResults = arts.getArtifact('roi')
if roiResults == None:
continue
print('Going over experiment: %s' % (filename, ))
currentExpResults = roiResults['arrivedFrac']
timePointResults = np.array(currentExpResults)[interestingTimePoints]
firstGroupResults[i, :] = timePointResults
print('Second Group:')
for i, filename in enumerate(secondExperimentsFiles):
if 'Nov' not in str(filename):
continue
if '03-Nov' in str(filename):
continue
# Getting the dirname of the experiment.
expDirName = path.dirname(filename)
arts = Artifacts(expLocation=expDirName)
roiResults = arts.getArtifact('roi')
if roiResults == None:
continue
print('Going over experiment: %s' % (filename, ))
currentExpResults = roiResults['arrivedFrac']
timePointResults = np.array(currentExpResults)[interestingTimePoints]
secoundGroupResults[i, :] = timePointResults
firstGroupResults = firstGroupResults[
np.logical_not(np.all(firstGroupResults == 0, axis=1)), :]
secoundGroupResults = secoundGroupResults[
np.logical_not(np.all(secoundGroupResults == 0, axis=1)), :]
flattenFirstGroup = firstGroupResults.flatten(order='C')
flattenSecondGroup = secoundGroupResults.flatten(order='C')
fracitons = tuple(flattenFirstGroup) + tuple(flattenSecondGroup)
conds = (firstGroupName, ) * flattenFirstGroup.size + (
secondGroupName, ) * flattenSecondGroup.size
timePoints = int((len(fracitons)) /
len(interestingTimePoints)) * tuple(interestingTimePoints)
df = pd.DataFrame({
'Fraction Arrived': fracitons,
'Cond': conds,
'time': timePoints
})
sns.set(style='darkgrid')
plt.style.use("dark_background")
ax = sns.pointplot(x='time', y='Fraction Arrived', data=df, hue='Cond')
ax.set(xlabel='Frames [2Hz]')
plt.show()
pairedDf_2 = pd.DataFrame({
'ATR+':
firstGroupResults[[1, 2, 3, 4, 5, 6, 7, 9], 2],
'ATR-':
secoundGroupResults[[2, 3, 4, 5, 6, 7, 8, 9], 2]
})
pairedDf_3 = pd.DataFrame({
'ATR+':
firstGroupResults[[1, 2, 3, 4, 5, 6, 7, 9], 3],
'ATR-':
secoundGroupResults[[2, 3, 4, 5, 6, 7, 8, 9], 3]
})
pairedDf_4 = pd.DataFrame({
'ATR+':
firstGroupResults[[1, 2, 3, 4, 5, 6, 7, 9], 4],
'ATR-':
secoundGroupResults[[2, 3, 4, 5, 6, 7, 8, 9], 4]
})
pairedDf = pd.concat((pairedDf_2, pairedDf_3, pairedDf_4),
ignore_index=True)
ax = sns.scatterplot(x='ATR-', y='ATR+', data=pairedDf)
x0, x1 = ax.get_xlim()
y0, y1 = ax.get_ylim()
lims = [max(x0, y0), min(x1, y1)]
ax.plot(lims, lims, ':k', color='white')
print('Done.')
def createDemMovie(exp_dir):
expDir = ExpDir(exp_dir)
artifacts = Artifacts(expLocation=exp_dir)
# Getting the frame intensities values
frame_intensities = artifacts.getArtifact('frame_intensities')
frame_intensities /= np.max(frame_intensities)
frame_range = range(1200, 3000)
#cap = cv2.VideoCapture(expDir.getVidFile())
cap = cv2.VideoCapture(expDir.getExpSegVid())
cap.set(cv2.CAP_PROP_POS_FRAMES, np.min(frame_range))
frame_intensities = frame_intensities[frame_range]
plt.style.use('dark_background')
plt.tight_layout()
sns.set_context('paper')
# Setting a 3 to 1 ratio between video and figure.
fig, axs = \
plt.subplots(2, 3, gridspec_kw={'height_ratios':[4, 1], 'width_ratios':[1 ,3, 1]})
ax_vid = axs[0][1]
ax_fig = axs[1][1]
ax_vid.axis('off')
axs[0][0].axis('off')
axs[0][2].axis('off')
axs[1][0].axis('off')
axs[1][2].axis('off')
# Global variables
im = None
vidAlign = 0
vertLine = None
def updateMovie(frameNumber):
nonlocal ax_vid
nonlocal ax_fig
nonlocal cap
nonlocal im
nonlocal vidAlign
nonlocal frame_range
nonlocal vertLine
print('Frame number: %d' % frameNumber)
#DEBUG
cap.set(cv2.CAP_PROP_POS_FRAMES, frameNumber)
#DEBUG
# Updating the video
fig.sca(ax_vid)
_, frame = cap.read()
frame = frame[480:1440, 1100:2300]
if frameNumber == -1:
vidAlign = alignImage(frame)
frame = np.roll(frame, vidAlign, axis=1)
im = plt.imshow(frame)
cap.set(cv2.CAP_PROP_POS_FRAMES, np.min(frame_range))
else:
frame = np.roll(frame, vidAlign, axis=1)
im.set_data(frame)
# Updating the figure
fig.sca(ax_fig)
if frameNumber == -1:
plt.gca().grid(alpha=0.2)
plt.plot(np.array(frame_range) * 0.5,
frame_intensities,
linewidth=1.5,
color=sns.xkcd_rgb['windows blue'],
alpha=0.75)
plt.xlabel('Time [s]')
h = plt.ylabel('Power [au]')
#h.set_rotation(0)
plt.xlim(np.min(frame_range) * 0.5, np.max(frame_range) * 0.5)
else:
if vertLine is not None:
vertLine.remove()
vertLine = ax_fig.axvline(x=frameNumber * 0.5, ymin=0, ymax=1)
return (ax_vid, ax_fig)
anim = FuncAnimation(fig,
updateMovie,
frames=[1905],
init_func=lambda: updateMovie(-1))
anim.save('/home/itskov/Dropbox/DemMovie2.mp4',
fps=50,
extra_args=['-vcodec', 'libx264'],
dpi=250)
cap.release()
def gatherTracksForSpikes(exp_file, show_plot=False):
RANGE_BEFORE = 100
RANGE_AFTER = 260
MAX_FRAMES = 3500
speed_up_times = []
exp = np.load(exp_file)[0]
# Filtering the tracks.
exp._tracks = filterTracksForAnalyses(exp._tracks,
minSteps=370,
minDistance=150)
# Get pulses in time
dir_name = path.dirname(exp._videoFilename)
#pulses_filename = path.join(dir_name, 'pulses.npy')
#print(pulses_filename)
artifacts = Artifacts(expLocation=dir_name)
artifacts.checkForArtifactsDir()
frame_intensities = artifacts.getArtifact('frame_intensities')
if frame_intensities is not None:
print('Found file.')
else:
frame_intensities = np.zeros((MAX_FRAMES, ))
for i in range(MAX_FRAMES - 1):
_, cur_frame = exp._cap.read()
frame_intensities[i] = np.mean(cur_frame)
print('Went over frame: %d' % i)
artifacts.addArtifact('frame_intensities', frame_intensities)
# Looking for spike points
spike_points = np.where(np.diff(frame_intensities) > 10)[0]
spike_points = spike_points[1:]
spike_ranges = []
for spike_point in spike_points:
spike_ranges.append(
list(range(spike_point - RANGE_BEFORE, spike_point + RANGE_AFTER)))
tracks_for_spikes = []
tracks_for_single_spikes = {}
df = pd.DataFrame({'frame': [], 'speed': []})
# Going over the relevant tacks.
for track in exp._tracks:
current_df = pd.DataFrame({
'frame': track._trackFrames,
'speed': track._tracksSpeeds
})
df = pd.concat((df, current_df))
for i, spike_range in enumerate(spike_ranges):
# DEBUG
# if i != 3:
# continue
# DEBUG
if spike_range[0] in track._trackFrames and spike_range[
-1] in track._trackFrames:
new_track = track.trimTrack(spike_range[-1], spike_range[0])
tracks_for_spikes.append(new_track)
tracks_for_single_spikes[i] = tracks_for_single_spikes[i] + [new_track]\
if i in tracks_for_single_spikes.keys() else [new_track]
pass
# Pick spike light represenetitive
spike_rep = frame_intensities[spike_ranges[1]]
spikes_speeds = np.array([t._tracksSpeeds for t in tracks_for_spikes])
spikes_angles = np.array([
t.getAngles(exp._regionsOfInterest['endReg']['pos'])
for t in tracks_for_spikes
])
spikes_reversals = np.array(
[t._tracksReversals for t in tracks_for_spikes])
spikes_angles[:, 0] = spikes_angles[:, 1]
spikes_angles[:, -2] = spikes_angles[:, -3]
spikes_angles[:, -1] = spikes_angles[:, -3]
mean_spikes_speeds = np.median(spikes_speeds, axis=0)
mean_spikes_angles = np.median(spikes_angles, axis=0)
mean_spikes_speeds = mean_spikes_speeds[2:-2]
spike_rep = spike_rep[2:-2]
# Gathering the speeding up time.
'''smoothed_speeds = savgol_filter(mean_spikes_speeds, 11, 3)
spike_decay_point = np.max(np.where(np.diff(spike_rep) < -5)) + 3
speed_up_time = \
np.min(np.where(smoothed_speeds[spike_decay_point:] >= np.mean(smoothed_speeds[0:RANGE_BEFORE])))
speed_up_times.append(speed_up_time)'''
artifacts.addArtifact('tracks_by_pulse', tracks_for_single_spikes)
artifacts.addArtifact('tracks_for_all_spikes', tracks_for_spikes)
def demonstrateLightingSegmentation(exp_dir):
expDir = ExpDir(exp_dir)
artifacts = Artifacts(expLocation=exp_dir)
# Getting the light intensity file
light_intensities = artifacts.getArtifact('frame_intensities')
base_line = np.mean(light_intensities[1:10])
#vid_cap = cv2.VideoCapture(expDir.getVidFile())
vid_cap = cv2.VideoCapture(
'/mnt/storageNASRe/tph1/12.03.20/12-Mar-2020-21.16.44-MIC2-TPH_1_ATR_ONLINE[NO_IAA]_5S.avi.mj2'
)
seg_cap = cv2.VideoCapture(expDir.getExpSegVid())
spike_count = 0
in_spike = False
in_spike_count = 0
full_vid_handle = FFmpegWriter('/home/itskov/Dropbox/dem.mp4',
outputdict={'-crf': '0'})
# Reading first frame for shape.
_, first_frame = vid_cap.read()
_, second_frame = seg_cap.read()
# Alignining only the segmented version.
seg_frame_align = alignImage(second_frame)
start_frame = 700
vid_cap.set(cv2.CAP_PROP_POS_FRAMES, start_frame)
seg_cap.set(cv2.CAP_PROP_POS_FRAMES, start_frame)
for i in range(start_frame, 2000):
_, vid_frame = vid_cap.read()
_, seg_frame = seg_cap.read()
vid_frame = cv2.cvtColor(vid_frame, cv2.COLOR_BGR2GRAY)
vid_frame = cv2.cvtColor(vid_frame, cv2.COLOR_GRAY2BGR)
#vid_frame = np.reshape(vid_frame, (vid_frame.shape))
vid_frame = np.roll(vid_frame, seg_frame_align, axis=1)
seg_frame = np.roll(seg_frame, seg_frame_align, axis=1)
if light_intensities[i] > (base_line + 7):
in_spike = True
in_spike_count += 1
if in_spike_count == 6 and spike_count % 2 == 0 and spike_count > 0:
splits_num = 80
width = vid_frame.shape[1]
segments = [
tuple(s)
for s in np.split(np.array(range(width)), splits_num)
]
for i in range(2 * splits_num - 1):
bar_step = i % (splits_num * 2)
if bar_step > splits_num:
bar_step = splits_num - (bar_step - splits_num)
if bar_step > 0 and bar_step < splits_num:
print(bar_step)
#print(segments)
#print(segments[bar_step:])
first_part = vid_frame[:,
np.ravel(segments[bar_step:]
), :]
second_part = seg_frame[:,
np.ravel(segments[0:bar_step]
), :]
new_frame = np.concatenate((second_part, first_part),
axis=1)
cur_img = Image.fromarray(new_frame)
cur_imgdraw = ImageDraw.Draw(cur_img)
cur_imgdraw.ellipse((40, 40, 180, 180),
fill='red',
outline='red')
full_vid_handle.writeFrame(np.array(cur_img))
else:
# Do Spike stuff here.
cur_img = Image.fromarray(vid_frame)
cur_imgdraw = ImageDraw.Draw(cur_img)
cur_imgdraw.ellipse((40, 40, 180, 180),
fill='red',
outline='red')
full_vid_handle.writeFrame(np.array(cur_img))
else:
if in_spike == True:
in_spike = False
spike_count += 1
in_spike_count = 0
full_vid_handle.writeFrame(vid_frame)
print('Frame %d. Spike Count; %d' % (i, spike_count))
vid_cap.release()
seg_cap.release()
full_vid_handle.close()
from Behavior.Visualizers.OccupVisualizer import OccupVisualizer
import numpy as np
rootDir = '/home/itskov/Temp/Chris/'
for fileName in Path(rootDir).rglob('*/exp.npy'):
print(fileName)
try:
# load experiment.
exp = np.load(fileName)[0]
exp.trimExperiment(4500)
# Create an artifact folder.
art = Artifacts(exp)
# Checking for artifact dirs.
art.checkForArtifactsDir()
roiAnalyses = RoiAnalysis(exp)
projectionAnalyses = ProjectionAnalyses(exp)
occupAnalyses = OccupVisualizer(exp)
roiAnalyses.execute()
projectionAnalyses.execute()
occupAnalyses.execute(showPlot=False)
#DEBUG Chris
exp._regionsOfInterest['startReg']['rad'] = exp._scale / 4
exp._regionsOfInterest['endReg']['rad'] = exp._scale / 3
def analyzeSpeedUps(exps, paper=False, fitLinear=True, show=True):
delays = []
speed_ups = []
for expDir, delay in exps:
artifacts = Artifacts(expLocation=expDir)
frame_intensities = artifacts.getArtifact('frame_intensities')
tracks_for_all_spikes = artifacts.getArtifact('tracks_for_all_spikes')
speeds = [t._tracksSpeeds for t in tracks_for_all_spikes]
signal = np.mean(speeds, axis=0)
signal = signal[2:-2]
min_decay_ind = np.argmin(signal)
signal = signal[min_decay_ind:]
signal = signal / signal[-1]
xdata = np.array(range(signal.shape[0])) * 0.5
popt, pcov = curve_fit(func, xdata, signal)
plt.plot(xdata, signal)
plt.plot(xdata, func(xdata, *popt), 'r-')
print(popt)
delays.append(delay)
speed_ups.append(popt[1])
pass
regressor = LinearRegression()
delays = np.array(delays).reshape(-1, 1)
speed_ups = np.array(speed_ups).reshape(-1, 1)
regressor.fit(delays, speed_ups)
predicted = regressor.predict(delays)
linearDf = pd.DataFrame({
'Activation Duration [s]': np.ravel(delays),
'Time Coefficient [s]': np.ravel(speed_ups),
'Predicted': np.ravel(predicted)
})
if not fitLinear:
print(np.ravel(delays))
print(np.ravel(speed_ups))
popt, pcov = curve_fit(func, np.ravel(delays), np.ravel(speed_ups))
plt.close()
if not paper:
sns.set_context('talk')
plt.style.use("dark_background")
cp = sns.dark_palette("purple", 7)
if fitLinear:
sns.lineplot(data=linearDf,
x='Activation Duration [s]',
y='Predicted',
ci=None,
color=cp[2])
else:
xvals = np.linspace(np.min(delays), np.max(delays), 1000)
plt.plot(xvals, func(xvals, *popt), color=cp[2])
sns.scatterplot(data=linearDf,
x='Activation Duration [s]',
y='Time Coefficient [s]',
linewidth=0,
alpha=0.98,
color=cp[6])
else:
sns.set_context('paper')
if fitLinear:
sns.lineplot(data=linearDf,
x='Activation Duration [s]',
y='Predicted',
ci=None)
else:
xvals = np.linspace(np.min(delays), np.max(delays), 1000)
plt.plot(xvals, func(xvals, *popt))
sns.scatterplot(data=linearDf,
x='Activation Duration [s]',
y='Time Coefficient [s]',
linewidth=0,
alpha=0.98)
plt.title('$R^2 = %.2f$' % (r2_score(speed_ups, predicted), ), loc='left')
plt.gca().grid(alpha=0.2)
if show:
plt.show()
pass