def setup(filelist, prestim_frames=400, **kwargs):
stores = []
framenums = []
for vid in filelist:
pf = pd.read_pickle(vid + '/track/perframe_stats.pickle')
log = stims.get_logfile(vid)
store = imgstore.new_for_filename(vid + '/metadata.yaml')
stores.append(store)
ret, pf2 = stims.synch_coherence_with_rotation(pf, log,
store) #sync_reversals
pf3 = sync_by_stimStart(pf2)
startframe = pf3.loc[pf3['stimStart'].idxmax() - prestim_frames,
'FrameNumber']
framenums.append(startframe)
return filelist, stores, framenums
def run(MAIN_DIR, RESUME=True):
print("processing: ", MAIN_DIR)
#getColumnNames('_'.join(MAIN_DIR.split('/')[-1]..split('.')[0].split('_')[-2:]))
trackdir = slashdir(MAIN_DIR) + 'track/'
PF_DONE = False
if os.path.exists(trackdir + 'perframe_stats.pickle'):
if datetime.datetime.fromtimestamp(
os.path.getmtime(trackdir + 'perframe_stats.pickle')
) > getTimeFromTimeString('20191218_000000'):
PF_DONE = True
perframe_stats = pd.read_pickle(trackdir + 'perframe_stats.pickle')
if len(perframe_stats.shape) == 1:
perframe_stats = joblib.load(trackdir +
'perframe_stats.pickle')
if len(perframe_stats.shape) == 1:
PF_DONE = False
if PF_DONE == False:
if os.path.exists(trackdir + 'frameByFrameData.pickle'):
try:
fbf = pd.read_pickle(trackdir + 'frameByFrameData.pickle')
except:
fbf = joblib.load(trackdir + 'frameByFrameData.pickle')
if len(fbf.shape) == 1:
fbf = getFrameByFrameData(trackdir, RESUME, args.maxthreads)
#print("CORRUPTED FILE. DELETING frameByFrameData:", trackdir)
#os.remove(trackdir + 'frameByFrameData.pickle')
#return
else:
fbf = getFrameByFrameData(trackdir, RESUME, args.maxthreads)
if not 'VX#smooth#wcentroid' in fbf.columns:
print('VX#smooth#wcentroid', "not found in columns.", MAIN_DIR)
print(fbf.columns)
os.remove(trackdir + 'frameByFrameData.pickle')
shutil.rmtree(trackdir + 'fishdata')
if os.path.exists(trackdir + 'frameByFrame_complete'):
os.remove(trackdir + 'frameByFrame_complete')
return
if len(set(fbf.frame)) < 501:
print("FOUND INCOMPLETE TRACKING DATA. DELETING TRACKDIR")
shutil.rmtree(trackdir)
return
perframe_stats = calculate_perframe_stats(fbf, trackdir,
args.maxthreads)
store = imgstore.new_for_filename(slashdir(MAIN_DIR) + 'metadata.yaml')
log = stim_handling.get_logfile(MAIN_DIR)
if 'reversals' in MAIN_DIR:
#ret, perframe_stats = stim_handling.sync_reversals(perframe_stats, log, store)
plot_perframe_vs_time(slashdir(MAIN_DIR), [
'dir', 'median_polarization', 'median_dRotation_cMass',
'median_dRotation_cArea', 'median_swimSpeed', 'entropy_Ra'
], [
'Direction', 'Pol. Order', 'Rot. Order (CofM)',
'Rot. Order (Area)', 'Median Speed', 'Entropy'
], perframe_stats, '_median')
elif 'coherence' in MAIN_DIR:
#ret, perframe_stats = stim_handling.synch_coherence_with_rotation(perframe_stats, log, store)
plot_perframe_vs_time(slashdir(MAIN_DIR), [
'coherence', 'median_polarization', 'median_dRotation_cMass',
'median_dRotation_cArea', 'median_swimSpeed', 'entropy_Ra'
], [
'Coherence', 'Pol. Order', 'Rot. Order (CofM)',
'Rot. Order (Area)', 'Median Speed', 'Entropy'
], perframe_stats, '_median')
elif 'cogs' in MAIN_DIR:
pass #FIXME
def calculate_perframe_stats(fbf, TRACK_DIR, nCores=8):
# SETUP PARALLEL PROCESSING
ppe = ProcessPoolExecutor(nCores)
futures = []
statResults = []
rotMResults = []
rotAResults = []
fishR = []
# PREPARE DATAFRAME
#fbf = fbf.loc[fbf['frame'] < 20100, :] #large files cause memory error on this old machine
fbf = fbf.loc[fbf[XPOS].notnull(), :]
fbf = fbf.loc[fbf[YPOS].notnull(), :]
fbf.loc[:, 'uVX'] = fbf.loc[:, XVEL] / fbf.loc[:, SPEED]
fbf.loc[:, 'uVY'] = fbf.loc[:, YVEL] / fbf.loc[:, SPEED]
if 'header' in fbf.columns:
fbf = fbf.drop(columns=['header'])
fbf['coreGroup'] = fbf[
'frame'] % nCores #divide into chunks labelled range(nCores)
fbf.reset_index(inplace=True, drop=True)
# INITIATE PARALLEL PROCESSES
for n in range(nCores):
p = ppe.submit(process_chunk, fbf.loc[fbf['coreGroup'] == n, :])
futures.append(p)
# COLLECT PROCESSED DATA AS IT IS FINISHED
for future in as_completed(futures):
stats, rotM, rotA, fish = future.result()
statResults.append(stats)
rotMResults.append(rotM)
rotAResults.append(rotA)
fishR.append(fish)
#CONCATENATE RESULTS
perframe_stats = pd.concat(statResults)
rotationDf = pd.concat(fishR)
rotationOrders_cMass = {}
for r in rotMResults:
rotationOrders_cMass.update(r)
pick = open(TRACK_DIR + '/rotationOrders_cMass.pickle', "wb")
pickle.dump(rotationOrders_cMass, pick)
pick.close()
rotationOrders_cArea = {}
for r in rotAResults:
rotationOrders_cArea.update(r)
pick = open(TRACK_DIR + '/rotationOrders_cArea.pickle', "wb")
pickle.dump(rotationOrders_cArea, pick)
pick.close()
ARENA_WIDTH = get_arena_width(TRACK_DIR.split('/track')[0])
#perframe_stats.loc[:,'centroidRotation'] = get_centroid_rotation(perframe_stats, TRACK_DIR, ARENA_WIDTH)
perframe_stats['frame'] = perframe_stats.index
log = stim_handling.get_logfile(TRACK_DIR.rsplit('/', 2)[0])
store = imgstore.new_for_filename(
TRACK_DIR.rsplit('/', 2)[0] + '/metadata.yaml')
ret, perframe_stats = stim_handling.sync_data(perframe_stats, log, store)
perframe_stats.to_pickle(TRACK_DIR + '/perframe_stats.pickle')
try:
rotationDf.to_pickle(TRACK_DIR + '/frameByFrameData.pickle')
except:
import joblib
joblib.dump(rotationDf, TRACK_DIR + '/frameByFrameData.jl')
return perframe_stats
import os
import pandas as pd
import numpy as np
MAIN_DIR = '/media/recnodes/recnode_2mfish/reversals3m_64_dotbot_20181024_153201.stitched/'
DATA_COL = 'EigenCen'
STIM_COL = 'dir'
SKIP_FRAMES = 3
SAVE_DIR = MAIN_DIR + 'track/eigencentricity_overlay/'
nfbf = joblib.load(MAIN_DIR + 'track/network_FBF.pickle')
nfbf.index = nfbf.frame
store = imgstore.new_for_filename(MAIN_DIR + 'metadata.yaml')
log = stims.get_logfile(MAIN_DIR)
ret, pf = stims.sync_data(
pd.read_pickle(MAIN_DIR + 'track/perframe_stats.pickle'), log, store)
ret, nfbf = stims.sync_data(nfbf, log, store)
frames = pd.DataFrame(store.get_frame_metadata())
frames.columns = ['FrameNumber', 'Timestamp']
_MIN = nfbf[DATA_COL].min()
_MAX = nfbf[DATA_COL].max()
g = nfbf.groupby('frame')
for i in np.arange(0, len(frames), SKIP_FRAMES):
if os.path.exists(SAVE_DIR + '%06d.png' % i):
def get_Tseries(expFileName):
expID = expFileName.split('/')[-1].split('_', 3)[-1].split('.')[0]
if 'reversal' in expFileName:
REVERSAL = True
prestim_frames = 400
poststim_frames = 2400
file_prefix = 'REV_'
COLUMN = 'dir'
else:
REVERSAL = False
prestim_frames = 1200
poststim_frames = 16000
file_prefix = 'COH_'
COLUMN = 'speed'
try:
fbf = pd.read_pickle(expFileName + '/track/frameByFrameData.pickle')
if len(fbf.shape) == 1:
fbf = joblib.load(expFileName + '/track/frameByFrameData.pickle')
pf = pd.read_pickle(
expFileName + '/track/perframe_stats.pickle'
) #FIXME these may have the wrong stim direction because of sync_data vs sync_coherence (if made before 20191218)....
if len(pf.shape) == 1:
pf = joblib.load(expFileName + '/track/perframe_stats.pickle')
if not 'frame' in pf.columns:
fbf['frame'] = pf.index
if not 'FrameNumber' in pf.columns:
try:
ret, pf = stims.sync_data(
pf, stims.get_logfile(expFileName),
imgstore.new_for_filename(expFileName + '/metadata.yaml'))
except:
print("oppala")
return 0
ret, sy = sync_by_stimStart(pf, expID, col=COLUMN, REVERSALS=REVERSAL)
if ret == 0:
return 0
except Exception as e:
print(e)
return 0
for ID, data in sy.groupby('trialID'):
frame_0 = data.loc[data['syncTime'] == abs(data.syncTime).min(),
'frame'].values[0]
md = data.loc[frame_0 - prestim_frames:frame_0 + poststim_frames]
prestim = data.loc[data['frame'] < frame_0, :]
psMeta = dict(prestim[prestim.columns[prestim.dtypes ==
'float64']].fillna(0).mean())
a = area[area.index == ID]
if len(a) > 0:
for col in [
'Area', 'Area_rank', 'Density', 'Density_rank',
'groupsize', 'trialID'
]:
psMeta[col] = a[col][0]
else:
psMeta['trialID'] = ID
d = fbf[fbf['frame'].between(frame_0 - prestim_frames,
frame_0 + poststim_frames)]
rotA = d.groupby(['frame',
'trackid'])['rotation_cArea'].mean().unstack()
rotM = d.groupby(['frame',
'trackid'])['rotation_cMass'].mean().unstack()
stimdir = md['dir'].fillna(0)
stimcoh = md['coh'].fillna(0)
stimspeed = md['speed'].fillna(0)
meta = {}
COH = str(np.around(pf.coh.median(), 1))
GS = expFileName.split('/')[-1].split('_')[1]
#ID = expFileName.split('/')[-1].split('_',3)[-1].split('.')[0]
FN = '/media/recnodes/Dan_storage/Jake_TS/' + file_prefix + GS + '_' + COH + '_' + ID + '.npz'
#FN = '/media/recnodes/Dan_storage/Jake_TS/'+ expFileName.split('/')[-1].rsplit('_',2)[0] + '_' + ID + '.npz'
np.savez(FN,
rA=rotA,
rM=rotM,
prestim=psMeta,
meta=meta,
direction=stimdir,
coherence=stimcoh,
speed=stimspeed)
print('completed:', FN)
return 1
def save_rotation_data(expFileName):
if 'reversal' in expFileName:
REVERSAL = True
XLIM = (30, 60)
COLUMN = 'dir'
else:
REVERSAL = False
XLIM = (30, 300)
COLUMN = 'speed'
try:
fbf = pd.read_pickle(expFileName + '/track/perframe_stats.pickle')
if len(fbf) == 0:
print("FAILED TO READ PICKLE. TRYING JOBLIB")
fbf = joblib.load(expFileName + '/track/perframe_stats.pickle')
rot = pd.read_pickle(expFileName +
'/track/rotationOrders_cArea.pickle')
if not 'frame' in fbf.columns:
fbf['frame'] = fbf.index
if not 'FrameNumber' in fbf.columns:
ret, fbf = stims.sync_data(
fbf, stims.get_logfile(expFileName),
imgstore.new_for_filename(expFileName + '/metadata.yaml'))
ID = expFileName.split('/')[-1].split('_', 3)[-1].split('.')[0]
ret, synced = sync_by_stimStart(fbf,
ID,
col=COLUMN,
REVERSALS=REVERSAL)
if ret == 0:
return 0
for IDx in list(set(synced['trialID'])):
chunk = synced.loc[synced['trialID'] == IDx, :]
ix = chunk[chunk.syncTime.between(np.timedelta64(XLIM[0], 's'),
np.timedelta64(XLIM[1],
's'))].index
DIR = np.sign(chunk.loc[ix, 'dir'].mean())
if DIR == 0:
return 0
data = np.concatenate([
rot[x] for x in range(ix.min(), ix.max())
]) * DIR * -1.0 #FLIP TO MAKE POSITIVE, JAKE
COH = str(np.around(fbf.coh.mean(), 1))
GS = expFileName.split('/')[-1].split('_')[1]
if REVERSAL:
np.save(
'/media/recnodes/Dan_storage/20200121_reversal_rotation_data/'
+ GS + '_' + COH + '_' + IDx + '.npy', data)
print(
'/media/recnodes/Dan_storage/20200121_reversal_rotation_data/'
+ GS + '_' + COH + '_' + IDx + '.npy')
else:
continue #FIXME danno's too chicken to try this cuz it'll surely break
np.save(
'/media/recnodes/Dan_storage/191205_rotation_data/' + GS +
'_' + COH + '_' + ID + '.npy', data)
print('/media/recnodes/Dan_storage/191205_rotation_data/' +
GS + '_' + COH + '_' + ID + '.npy')
return 1
except Exception as e:
print(e)
return 0
if not os.path.exists(MAIN_DIR + '/track/localData_FBF.pickle'):
continue
expID = MAIN_DIR.split('/')[-1].split('.')[0]
exp, groupsize, _, trialID = expID.split('_', 3)
#don't repeat if already done
if trialID in results.index:
continue
try:
store = imgstore.new_for_filename(MAIN_DIR + '/metadata.yaml')
fbf = pd.read_pickle(MAIN_DIR + '/track/frameByFrameData.pickle')
if len(fbf) < 1000:
fbf = joblib.load(MAIN_DIR + '/track/frameByFrameData.pickle')
ret, fbf = stims.sync_data(fbf, stims.get_logfile(MAIN_DIR), store)
fbf = drop_bad_points(fbf)
if 'coherence' in expID:
synced = sync_by_stimStart(fbf, trialID)
elif 'reversal' in expID:
synced = sync_by_reversal(fbf, trialID)
prestim = synced[synced.syncTime.between(np.timedelta64(-30, 's'),
np.timedelta64(0, 's'))]
for group, data in prestim.groupby('trialID'):
df = get_prestim_area(data)
df['exp'] = exp
df['groupsize'] = groupsize
s = pd.Series(df.median(), name=group)
lastStim = df.loc[df['Time'] > df['Time'].median(), 'speed'].idxmin()
df.loc[lastStim, 'stimEnd'] = 1
return df
allData = pd.DataFrame()
for fn in glob.glob(
'/media/recnodes/recnode_2mfish/coherencetestangular3m_*_dotbot_*/track/perframe_stats.pickle'
):
expID, groupsize, _, trialID = fn.split('/')[4].split('.')[0].split('_', 3)
if fn.split('/track/perframe_stats')[0] in blacklist:
print "excluding", fn
continue
print fn
ret, pf = stims.sync_data(
pd.read_pickle(fn), stims.get_logfile(fn.rsplit('/', 2)[0]),
imgstore.new_for_filename(fn.rsplit('/', 2)[0] + '/metadata.yaml'))
pf['dir'] = pd.to_numeric(pf['dir'], errors='coerce')
pf['coh'] = pf['coh'].fillna(method='pad').fillna(method='backfill')
try:
pf = sync_by_stimStart(pf)
pf = align_by_stim(pf, trialID)
#slope = pd.Series(np.gradient(pf['median_dRotation_cArea'].values), pf['Timestamp'], name='slope')
s = splrep(pf.Timestamp, pf.median_dRotation_cArea, k=5, s=17)
newdf = pd.DataFrame({
'syncTime': pf['syncTime'],
'Orotation': pf['median_dRotation_cArea'],
'smoothedOrotation': splev(pf.Timestamp, s),
'dO_by_dt': splev(pf.Timestamp, s, der=1),
'dO_by_dt2': splev(pf.Timestamp, s, der=2)
def get_trial_data(MAIN_DIR):
try:
fbf = pd.read_pickle(MAIN_DIR + 'track/frameByFrameData.pickle')
assert len(fbf.shape) > 1
except:
fbf = joblib.load(MAIN_DIR + 'track/frameByFrameData.pickle')
assert len(fbf.shape) > 1
fbf['trackid'] = fbf['trackid'].astype(int)
try:
l = pd.read_pickle(MAIN_DIR + 'track/localData_FBF.pickle')
assert len(l.shape) > 1
except:
l = joblib.load(MAIN_DIR + 'track/localData_FBF.pickle')
assert len(l.shape) > 1
l['trackid'] = l['trackid'].astype(int)
l['frame'] = l['frame'].astype(int)
df = pd.merge(fbf,
l,
how='left',
left_on=['frame', 'trackid'],
right_on=['frame', 'trackid'])
store = imgstore.new_for_filename(MAIN_DIR + 'metadata.yaml')
log = stims.get_logfile(MAIN_DIR)
ret, synced = stims.sync_data(df, log, store)
#calculate individual rotation data:
synced['CX'] = synced[XPOS] - 160
synced['CY'] = synced[YPOS] - 167
synced['radius'] = np.sqrt(synced['CX']**2 + synced['CY']**2)
synced['uCX'] = synced['CX'] / synced['radius']
synced['uCY'] = synced['CY'] / synced['radius']
synced['uVX'] = synced[XVEL] / synced[SPEED]
synced['uVY'] = synced[YVEL] / synced[SPEED]
synced['Rotation'] = np.cross(synced[['uCX', 'uCY']],
synced[['uVX', 'uVY']])
synced = synced.sort_values('Timestamp').reset_index()
synced = synced[:-10]
synced['reversal'] = 0
reversals = synced[abs(synced['dir'] - synced.shift()['dir']) == 2].index
synced.loc[reversals, 'reversal'] = 1
synced.loc[synced['Time'] > 300,
'reversal'] = 0 #FIXME this is a hack solution to sort out ends
synced['firstStim'] = 0
firstStim = synced[synced['Time'] < synced['Time'].median()]
firstStim = firstStim[abs(firstStim['dir'] -
firstStim.shift()['dir']) == 1].index
synced.loc[firstStim, 'firstStim'] = 1
synced['lastStim'] = 0
lastStim = synced[synced['Time'] > synced['Time'].median()]
lastStim = lastStim[abs(lastStim['dir'] -
lastStim.shift()['dir']) == 1].index
synced.loc[lastStim, 'lastStim'] = 1
alignPoints = list(synced[synced['reversal'] == 1]['Timestamp'].values)
synced = synced[synced['dir'].isnull() == False]
trials = pd.DataFrame()
fileID = MAIN_DIR.split('/')[-2].split('.')[0].split('_', 3)[-1]
trialID = 0
for i in alignPoints:
data = synced.loc[synced['Timestamp'].between(i - 10.0, i + 60.0), [
'Timestamp', 'speed', 'dir', 'coh', 'frame', 'neighbourDist',
'localArea', 'localPackingFraction', 'localMedianRotation',
'localRscore', 'localPolarization', 'localPDcor',
'localSpeedScore', 'radius', 'Rotation'
]]
data['syncTime'] = pd.to_timedelta(data['Timestamp'] - i, 's')
data['localMedianRotation'] = data['localMedianRotation'] * np.sign(
data['dir'].median()) * -1.0 #make congruent and positive
data['Rotation'] = data['Rotation'] * np.sign(
data['dir'].median()) * -1.0 #make congruent and positive
data.index = data['syncTime']
data['syncTime'] = data.index.copy()
GBF = data.resample('250ms')
median = GBF.median()
median['Spearman_Rotation_neighbourDist'] = [
spearmanr(x['Rotation'], x['neighbourDist'])[0] for i, x in GBF
]
median['Spearman_Rotation_localArea'] = [
spearmanr(x['Rotation'], x['localArea'])[0] for i, x in GBF
]
median['Spearman_Rotation_localPackingFraction'] = [
spearmanr(x['Rotation'], x['localPackingFraction'])[0]
for i, x in GBF
]
median['Spearman_Rotation_localMedianRotation'] = [
spearmanr(x['Rotation'], x['localMedianRotation'])[0]
for i, x in GBF
]
median['Spearman_Rotation_localPolarization'] = [
spearmanr(x['Rotation'], x['localPolarization'])[0] for i, x in GBF
]
median['Spearman_Rotation_localPDcor'] = [
spearmanr(x['Rotation'], x['localPDcor'])[0] for i, x in GBF
]
median['Spearman_Rotation_localSpeedScore'] = [
spearmanr(x['Rotation'], x['localSpeedScore'])[0] for i, x in GBF
]
median['Spearman_Rotation_radius'] = [
spearmanr(x['Rotation'], x['radius'])[0] for i, x in GBF
]
median['trialID'] = fileID + '_' + str(trialID)
median['date'] = fileID.split('_')[0]
trialID += 1
trials = pd.concat([trials, median], axis=0)
return trials
groupsizes = [64,128,256,512,1024]
coherences = [0,0.2,0.4,0.6,0.8,1]
allData = pd.DataFrame()
prestim_meanvals = pd.DataFrame()
meanvals = pd.DataFrame()
for groupsize in groupsizes:
print groupsize
groupData = pd.DataFrame()
for fn in glob.glob('/media/recnodes/recnode_2mfish/coherencetestangular3m_' + str(groupsize) + '_dotbot_*/track/perframe_stats.pickle'):
if fn.split('/track/perframe_stats')[0] in blacklist:
print "excluding", fn
continue
print fn
ret, pf = stims.sync_data(pd.read_pickle(fn), stims.get_logfile(fn.rsplit('/',2)[0]), imgstore.new_for_filename(fn.rsplit('/',2)[0] + '/metadata.yaml'))
pf['dir'] = pd.to_numeric(pf['dir'], errors='coerce')
pf['coh'] = pf['coh'].fillna(method='pad').fillna(method='backfill')
prestim_mean = pf.loc[pf['Time'] < pf[pf['speed'] !=0]['Time'].min(), :].mean()
prestim_mean['ID'] = fn.split('/')[-3].split('_',3)[-1].split('.')[0]
prestim_mean['groupsize'] = groupsize
prestim_meanvals = pd.concat([prestim_meanvals, prestim_mean], axis=1)
_mean = pf.mean()
_mean['ID'] = fn.split('/')[-3].split('_',3)[-1].split('.')[0]
_mean['groupsize'] = groupsize
meanvals = pd.concat([meanvals, _mean], axis=1)
pf = sync_by_stimStart(pf)
fileID = fn.split('/')[-3].split('.')[0].split('_',3)[-1]
aligned = align_by_stim(pf, fileID)
aligned['groupsize'] = groupsize
for groupsize in groupsizes:
print groupsize
groupData = pd.DataFrame()
for fn in glob.glob('/media/recnodes/recnode_2mfish/reversals3m_' +
str(groupsize) +
'_dotbot_*/track/perframe_stats.pickle'):
if '20181026' in fn: #
continue
print fn
try:
pf = pd.read_pickle(fn)
if not 'FrameNumber' in pf.columns:
if not 'frame' in pf.columns:
pf['frame'] = pf.index.copy()
ret, pf = stims.sync_reversals(
pf, stims.get_logfile(fn.rsplit('/', 2)[0]),
imgstore.new_for_filename(
fn.rsplit('/', 2)[0] + '/metadata.yaml'))
except Exception as e:
print "failed for ", fn
print e
continue
pf['dir'] = pd.to_numeric(pf['dir'], errors='coerce')
pf = sync_by_reversal(pf)
fileID = fn.split('/')[-3].split('.')[0].split('_', 3)[-1]
reversals = align_by_stim(pf, fileID)
if len(reversals) == 0:
continue
reversals['groupsize'] = groupsize
"""
def doit(MAIN_DIR, saveas="Not_defined", nCores=16, window='undefined'):
# SETUP PARALLEL PROCESSING
ppe = ProcessPoolExecutor(nCores)
futures = []
Results = []
MAIN_DIR = slashdir(
MAIN_DIR) #sometimes people call dirs without a trailing slash
print "Processing: ", MAIN_DIR
if saveas == 'Not_defined':
saveas = MAIN_DIR + 'voronoi_overlay'
if not os.path.exists(saveas):
os.makedirs(saveas)
fbf = joblib.load(MAIN_DIR + 'track/frameByFrameData.pickle')
fbf = fbf.replace(to_replace=np.inf, value=np.nan)
# PREPARE DATAFRAME
fbf = fbf.loc[fbf[XPOS].notnull(), :]
fbf = fbf.loc[fbf[YPOS].notnull(), :]
fbf.loc[:, 'uVX'] = fbf.loc[:, XVEL] / fbf.loc[:, SPEED]
fbf.loc[:, 'uVY'] = fbf.loc[:, YVEL] / fbf.loc[:, SPEED]
#filter out tracked reflections by removing tracks that are always near the border
foo = fbf.groupby('trackid').max()['BORDER_DISTANCE#wcentroid']
fbf = fbf[~(fbf.trackid.isin(foo[foo < 50].index))]
arena_centre = (160.0, 160.0)
fbf['radius'] = np.sqrt((fbf[XPOS] - arena_centre[0])**2 +
(fbf[YPOS] - arena_centre[1])**2)
fbf = fbf.loc[fbf['radius'] < 215, :]
fbf = fbf.loc[fbf[XPOS] < 315, :] #FIXME hardcoding bad
if 'header' in fbf.columns:
fbf = fbf.drop(columns=['header'])
fbf['coreGroup'] = fbf[
'frame'] % nCores #divide into chunks labelled range(nCores)
store = imgstore.new_for_filename(MAIN_DIR + 'metadata.yaml')
if not 'FrameNumber' in fbf.columns:
ret, fbf = stim_handling.sync_data(fbf,
stim_handling.get_logfile(MAIN_DIR),
store)
fbf = sync_by_stimStart(fbf, MAIN_DIR.split('/')[-2].split('_', 3)[-1])
if not window == 'undefined':
fbf = fbf.loc[fbf.syncTime.between(np.timedelta64(window[0], 's'),
np.timedelta64(window[1], 's')), :]
# INITIATE PARALLEL PROCESSES
for n in range(nCores):
p = ppe.submit(process_chunk, fbf[fbf['coreGroup'] == n])
futures.append(p)
# COLLECT PROCESSED DATA AS IT IS FINISHED
for future in as_completed(futures):
stats = future.result()
Results.append(stats)
NN = pd.concat(Results)
if not window == 'undefined':
joblib.dump(
NN, MAIN_DIR + 'track/nearest_neighbours_FBF_' + str(window[0]) +
'-' + str(window[1]) + '.pickle')
else:
joblib.dump(NN, MAIN_DIR + 'track/nearest_neighbours_FBF.pickle')
return
]
return df.merge(foo, left_on='community', right_on='community'), edge2cid
from overlay_data_on_position import plot_data_on_video
from mpl_toolkits.axes_grid1 import make_axes_locatable
#MAIN_DIR = '/media/recnodes/recnode_2mfish/reversals3m_128_dotbot_20181211_151201.stitched/'
MAIN_DIR = '/media/recnodes/recnode_2mfish/reversals3m_128_dotbot_20181106_145202.stitched/'
store = imgstore.new_for_filename(MAIN_DIR + 'metadata.yaml')
from local_properties import sync_rotation
import stim_handling as stims
local = pd.read_pickle(MAIN_DIR + 'track/localData_FBF.pickle')
local = sync_rotation(local, stims.get_logfile(MAIN_DIR), store)
local['trackid'] = local['trackid'].astype(float).astype(int)
local['Time'] = local.loc[:, 'Timestamp'] - local.iloc[0]['Timestamp']
cols = [
'objIDcol', 'EigenCen', 'comm_EigenCen', 'localArea', 'comm_R', 'R',
'localMedianRotation', 'localRscore', 'localPackingFraction',
'localPolarization', 'localPScore', 'localSpeedScore'
]
cMin = [0, 0, 0.0, 0, 0, -1.0, -1.0, -1.0, 0, 0, 0, 0, 0]
cMax = [12, 0.2, 0.2, 6000, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
CT = CentroidTracker()
FBF = pd.DataFrame()
