def calc_dtheta(df):
#naieve method for calculating dtheta assuming wraparound.
dtheta = []
dtheta_idx = []
for i0,i1 in madplot.pairwise(df.iterrows()):
t0idx,t0row = i0
t1idx,t1row = i1
t0 = t0row['theta']
t1 = t1row['theta']
dt = (t1-t0)
#dtheta ranges from -pi/2.0 -> pi/2.0
dt = wrap_dtheta_plus_minus(dt, around=np.pi/2.0)
dtheta.append(dt)
dtheta_idx.append(t0idx)
return pd.Series(dtheta,index=dtheta_idx)
def prepare_data(path, resample_bin, gts):
LASER_THORAX_MAP = {True: THORAX, False: HEAD}
#PROCESS SCORE FILES:
pooldf = pd.DataFrame()
for df, metadata in flymad_analysis.load_courtship_csv(path):
csvfilefn, experimentID, date, time, genotype, laser, repID = metadata
dlaser = np.gradient(df['laser_state'].values)
num_on_periods = (dlaser == 0.5).sum()
if num_on_periods != 12:
print "\tskipping file %s (%d laser on periods)" % (
csvfilefn, num_on_periods / 2)
continue
if genotype not in gts:
print "\tskipping genotype", genotype
continue
duration = (df.index[-1] - df.index[0]).total_seconds()
if duration < EXPERIMENT_DURATION:
print "\tmissing data", csvfilefn
continue
print "\t%ss experiment" % duration
#make new columns that indicates HEAD/THORAX targeting
thorax = True
laser_state = False
trg = []
for i0, i1 in madplot.pairwise(df.iterrows()):
t0idx, t0row = i0
t1idx, t1row = i1
if t1row['laser_state'] >= 0.5 and t0row['laser_state'] == 0:
thorax ^= True
laser_state = True
elif t0row['laser_state'] >= 0.5 and t1row['laser_state'] == 0:
laser_state = False
trg.append(OFF if not laser_state else LASER_THORAX_MAP[thorax])
trg.append(OFF)
df['ttm'] = trg
#resample into 5S bins
df = df.resample(resample_bin, fill_method='ffill')
#trim dataframe
df = df.head(
flymad_analysis.get_num_rows(EXPERIMENT_DURATION, resample_bin))
tb = flymad_analysis.get_resampled_timebase(EXPERIMENT_DURATION,
resample_bin)
#fix cols due to resampling
df['laser_state'][df['laser_state'] > 0] = 1
df['zx_binary'] = (df['zx'] > 0).values.astype(float)
df['ttm'][df['ttm'] < 0] = HEAD
df['ttm'][df['ttm'] > 0] = THORAX
dlaser = np.gradient((df['laser_state'].values > 0).astype(int)) > 0
t0idx = np.argmax(dlaser)
t0 = tb[t0idx - 1]
df['t'] = tb - t0
#groupby on float times is slow. make a special align column
df['t_align'] = np.array(range(0, len(df))) - t0idx
df['obj_id'] = flymad_analysis.create_object_id(date, time)
df['Genotype'] = genotype
df['lasergroup'] = laser
df['RepID'] = repID
pooldf = pd.concat([pooldf, df])
data = {}
for gt in gts:
gtdf = pooldf[pooldf['Genotype'] == gt]
lgs = gtdf['lasergroup'].unique()
if len(lgs) != 1:
raise Exception("only one lasergroup handled for gt %s: not %s" %
(gt, lgs))
grouped = gtdf.groupby(['t'], as_index=False)
data[gt] = dict(mean=grouped.mean().astype(float),
std=grouped.std().astype(float),
n=grouped.count().astype(float),
first=grouped.first(),
df=gtdf)
return data
wfmf.show_timestamp = args.show_timestamp
wfmf.show_epoch = args.show_epoch
fmfwidth = wfmf.width
fmfheight = wfmf.height
try:
print 'loading w timestamps'
wts = wfmf.fmf.get_all_timestamps()
except AttributeError:
wts = df['t_framenumber'].dropna().unique()
pbar = madplot.get_progress_bar("computing frames", len(wts))
frames = []
for i, (wt0, wt1) in enumerate(madplot.pairwise(wts)):
pbar.update(i)
fdf = madplot.get_framedf(df, wt1)
lfdf = len(fdf)
if lfdf:
try:
most_recent_time = madplot.get_framedf(df, wt1).index[-1]
minidf = df[madplot.get_framedf(df, wt0).
index[0]:most_recent_time]
last_head_row = minidf[minidf['h_framenumber'].notnull()].tail(
1)
def prepare_data(path, resample_bin, gts):
LASER_THORAX_MAP = {True:THORAX,False:HEAD}
#PROCESS SCORE FILES:
pooldf = pd.DataFrame()
for df,metadata in flymad_analysis.load_courtship_csv(path):
csvfilefn,experimentID,date,time,genotype,laser,repID = metadata
dlaser = np.gradient(df['laser_state'].values)
num_on_periods = (dlaser == 0.5).sum()
if num_on_periods != 12:
print "\tskipping file %s (%d laser on periods)" % (csvfilefn, num_on_periods/2)
continue
if genotype not in gts:
print "\tskipping genotype", genotype
continue
duration = (df.index[-1] - df.index[0]).total_seconds()
if duration < EXPERIMENT_DURATION:
print "\tmissing data", csvfilefn
continue
print "\t%ss experiment" % duration
#make new columns that indicates HEAD/THORAX targeting
thorax = True
laser_state = False
trg = []
for i0,i1 in madplot.pairwise(df.iterrows()):
t0idx,t0row = i0
t1idx,t1row = i1
if t1row['laser_state'] >= 0.5 and t0row['laser_state'] == 0:
thorax ^= True
laser_state = True
elif t0row['laser_state'] >= 0.5 and t1row['laser_state'] == 0:
laser_state = False
trg.append(OFF if not laser_state else LASER_THORAX_MAP[thorax])
trg.append(OFF)
df['ttm'] = trg
#resample into 5S bins
df = df.resample(resample_bin, fill_method='ffill')
#trim dataframe
df = df.head(flymad_analysis.get_num_rows(EXPERIMENT_DURATION, resample_bin))
tb = flymad_analysis.get_resampled_timebase(EXPERIMENT_DURATION, resample_bin)
#fix cols due to resampling
df['laser_state'][df['laser_state'] > 0] = 1
df['zx_binary'] = (df['zx'] > 0).values.astype(float)
df['ttm'][df['ttm'] < 0] = HEAD
df['ttm'][df['ttm'] > 0] = THORAX
dlaser = np.gradient( (df['laser_state'].values > 0).astype(int) ) > 0
t0idx = np.argmax(dlaser)
t0 = tb[t0idx-1]
df['t'] = tb - t0
#groupby on float times is slow. make a special align column
df['t_align'] = np.array(range(0,len(df))) - t0idx
df['obj_id'] = flymad_analysis.create_object_id(date,time)
df['Genotype'] = genotype
df['lasergroup'] = laser
df['RepID'] = repID
pooldf = pd.concat([pooldf, df])
data = {}
for gt in gts:
gtdf = pooldf[pooldf['Genotype'] == gt]
lgs = gtdf['lasergroup'].unique()
if len(lgs) != 1:
raise Exception("only one lasergroup handled for gt %s: not %s" % (
gt, lgs))
grouped = gtdf.groupby(['t'], as_index=False)
data[gt] = dict(mean=grouped.mean().astype(float),
std=grouped.std().astype(float),
n=grouped.count().astype(float),
first=grouped.first(),
df=gtdf)
return data
