def prepare_data(bags, arena, smoothstr, smooth, medfilt, gts,
min_experiment_duration):
RESAMPLE_SPEC = '10L'
found_gts = []
pooldf = DataFrame()
for bag in bags:
df = madplot.load_bagfile_single_dataframe(bag,
arena,
ffill=True,
smooth=smooth)
df = flymad_analysis.resample(df, resample_specifier=RESAMPLE_SPEC)
metadata = filename_regexes.parse_filename(
bag, extract_genotype_and_laser=True)
dateobj = filename_regexes.parse_date(bag)
genotype = metadata['genotype'] + '-' + metadata['laser']
found_gts.append(genotype)
if genotype not in gts:
print "\tskipping genotype", genotype
continue
duration = (df.index[-1] - df.index[0]).total_seconds()
if duration < min_experiment_duration:
print "\tmissing data", bag
continue
print "\t%ss experiment" % duration
#MAXIMUM SPEED = 300:
#df['v'][df['v'] >= 300] = np.nan
#df['v'] = df['v'].fillna(method='ffill')
try:
df = flymad_analysis.align_t_by_laser_on(
df,
min_experiment_duration=min_experiment_duration,
align_first_only=True,
exact_num_ranges=1,
resample_bin=RESAMPLE_SPEC)
except flymad_analysis.AlignError, err:
print "\talign error %s (%s)" % (bag, err)
pass
#median filter
if medfilt:
df['v'] = scipy.signal.medfilt(df['v'].values, medfilt)
df['obj_id'] = calendar.timegm(dateobj)
df['Genotype'] = genotype
pooldf = pd.concat([pooldf, df])
def prepare_data(bags, arena, smoothstr, smooth, medfilt, gts, min_experiment_duration):
RESAMPLE_SPEC = '10L'
found_gts = []
pooldf = DataFrame()
for bag in bags:
df = madplot.load_bagfile_single_dataframe(bag, arena,
ffill=True,
smooth=smooth)
df = flymad_analysis.resample(df, resample_specifier=RESAMPLE_SPEC)
metadata = filename_regexes.parse_filename(bag, extract_genotype_and_laser=True)
dateobj = filename_regexes.parse_date(bag)
genotype = metadata['genotype'] + '-' + metadata['laser']
found_gts.append(genotype)
if genotype not in gts:
print "\tskipping genotype", genotype
continue
duration = (df.index[-1] - df.index[0]).total_seconds()
if duration < min_experiment_duration:
print "\tmissing data", bag
continue
print "\t%ss experiment" % duration
#MAXIMUM SPEED = 300:
#df['v'][df['v'] >= 300] = np.nan
#df['v'] = df['v'].fillna(method='ffill')
try:
df = flymad_analysis.align_t_by_laser_on(
df, min_experiment_duration=min_experiment_duration,
align_first_only=True,
exact_num_ranges=1,
resample_bin=RESAMPLE_SPEC)
except flymad_analysis.AlignError, err:
print "\talign error %s (%s)" % (bag, err)
pass
#median filter
if medfilt:
df['v'] = scipy.signal.medfilt(df['v'].values, medfilt)
df['obj_id'] = calendar.timegm(dateobj)
df['Genotype'] = genotype
pooldf = pd.concat([pooldf, df])
parser.add_argument('--outdir', help='destination directory for mp4')
args = parser.parse_args()
BAG_FILE = args.path[0]
ZOOM_FMF = args.zoom_fmf
WIDE_FMF = args.wide_fmf
wfmf = madplot.FMFTrajectoryPlotter(WIDE_FMF)
zfmf = madplot.FMFTTLPlotter(ZOOM_FMF)
zfmf.enable_color_correction(brightness=20, contrast=1.5)
arena = madplot.Arena(False)
print "loading data"
df = madplot.load_bagfile_single_dataframe(BAG_FILE, arena, ffill=True)
wt = wfmf.fmf.get_all_timestamps()
zt = zfmf.fmf.get_all_timestamps()
if len(df['tobj_id'].dropna().unique()) != 1:
print "TTM movies require single unique object IDs, I think..."
sys.exit(1)
frames = build_framedesc_list(df, wt, zt)
if not frames:
print "no frames to render"
sys.exit(0)
else:
print len(frames), "frames to render"
moviet0 = frames[0].epoch
help='show epoch')
parser.add_argument('--show-arena',
action='store_true',
default=False,
help='show arena')
parser.add_argument('--fps', type=int, default=20, help='framenumber')
parser.add_argument('--hist',
type=int,
default=400,
help='show this many points of trajectory history')
args = parser.parse_args()
path = args.path[0]
arena = madplot.Arena(False)
df = madplot.load_bagfile_single_dataframe(path, arena, ffill=False)
objids = df['tobj_id'].dropna().unique()
wfmf = madplot.FMFMultiTrajectoryPlotter(args.wide_fmf,
objids,
maxlen=args.hist)
if args.show_arena:
wfmf.show_arena = arena
wfmf.show_lxly = args.show_laser
wfmf.show_fxfy = args.show_target
wfmf.show_timestamp = args.show_timestamp
wfmf.show_epoch = args.show_epoch
fmfwidth = wfmf.width
fmfheight = wfmf.height
def prepare_data(arena, path):
pool_df = madplot.load_bagfile_single_dataframe(path,
arena,
ffill=False,
filter_short_pct=10.0)
targets = pool_df['target_type'].dropna().unique()
if len(targets) != 1:
raise Exception("Only head or body may be targeted")
if targets[0] == 1:
target_name = "head"
elif targets[0] == 2:
target_name = "body"
else:
raise Exception("Only head or body may be targeted")
ldf_all = pool_df[~pool_df['lobj_id'].isnull()]
hdf_all = pool_df[~pool_df['h_framenumber'].isnull()]
#the first fly targeted
prev = pool_df['lobj_id'].dropna().head(1)
prev_id = prev.values[0]
prev_ix = prev.index[0]
target_ranges = {
(0, 100): [],
}
for ix, row in pool_df.iterrows():
lobj_id = row['lobj_id']
if np.isnan(lobj_id):
continue
elif lobj_id != prev_id:
df = pool_df[prev_ix:ix]
#mean speed of fly over its whole turn
v = df['v'].mean()
#details of the TrackedObj message
ldf = df[~df['lobj_id'].isnull()]
#we want to know when we go to TTM mode
gbs = ldf.groupby('mode')
try:
#this is the time we first switch to TTM
to_ttm = gbs.groups[2.0][0]
print "------------", v
ttm_errs = []
ttm_dts_thisfly = []
ttm_dts = []
wf_errs = []
wf_dts_thisfly = []
wf_dts = []
#get the head detection data once we start ttm
hdf = pool_df[to_ttm:ix]
#extract only the head messages
for i, (hix, hrow) in enumerate(
hdf[~hdf['h_framenumber'].isnull()].iterrows()):
#skip when we miss the head
if hrow['head_x'] == 1e6:
continue
if target_name == "head":
ttm_err = math.sqrt( (hrow['head_x']-hrow['target_x'])**2 +\
(hrow['head_y']-hrow['target_y'])**2)
elif target_name == "body":
ttm_err = math.sqrt( (hrow['body_x']-hrow['target_x'])**2 +\
(hrow['body_y']-hrow['target_y'])**2)
#if the first value is < 10 it is likely actually associated
#with the last fly TTM, so ignore it....
if (i == 0) and (ttm_err < 100):
continue
#the time since we switched to this fly
thisfly_dt = (hix - prev_ix).total_seconds()
#the time since we switched to TTM
ttm_dt = (hix - to_ttm).total_seconds()
ttm_errs.append(ttm_err)
ttm_dts_thisfly.append(thisfly_dt)
ttm_dts.append(ttm_dt)
wdf = ldf[:hix].tail(1)
wf_err = math.sqrt( (wdf['fly_x']-wdf['laser_x'])**2 +\
(wdf['fly_y']-wdf['laser_y'])**2)
#the time since we switched to this fly
thisfly_dt = (wdf.index[0] - prev_ix).total_seconds()
#the time since we switched to TTM
ttm_dt = (wdf.index[0] - to_ttm).total_seconds()
wf_errs.append(wf_err)
wf_dts_thisfly.append(thisfly_dt)
wf_dts.append(ttm_dt)
print thisfly_dt, ttm_dt, ttm_err
ttm_err = Err(ttm_errs, ttm_dts_thisfly, ttm_dts)
wf_err = Err(wf_errs, wf_dts_thisfly, wf_dts)
trg = Target(lobj_id, prev_ix, ix, v, ttm_err, wf_err)
except KeyError:
#never switched to TTM
print "never switched to ttm"
trg = None
pass
#except Exception:
# print "UNKNOWN ERROR"
# trg = None
#classify the target into which speed range
if trg is not None:
for k in target_ranges:
vmin, vmax = k
if vmin < trg.v <= vmax:
target_ranges[k].append(trg)
prev_id = lobj_id
prev_ix = ix
data = {
'target_name': target_name,
'pooldf': pool_df,
'ldf': ldf_all,
'hdf': hdf_all,
'target_ranges': target_ranges
}
pkl_fname = "%s_%s.pkl" % (path, target_name)
cPickle.dump(data, open(pkl_fname, 'wb'), -1)
return data
def doit_using_framenumber(user_data):
match, mp4_dir, show_theta, show_velocity, pre_frames, post_frames = user_data
zoomf = match.fmf
rosbagf = match.bag
maxt = match.maxt
moviemaker = madplot.MovieMaker(basename=os.path.basename(zoomf), fps=15)
target_moviefname = moviemaker.get_target_movie_name(mp4_dir)
if os.path.exists(target_moviefname):
print 'target %r exists: skipping movie'%(target_moviefname,)
moviemaker.cleanup()
return
arena = madplot.Arena(False)
zoom = madplot.FMFImagePlotter(zoomf, 'z_frame')
zoom.enable_color_correction(brightness=15, contrast=1.5)
wide = madplot.ArenaPlotter(arena,tzname=TZNAME)
wide.show_theta = show_theta
wide.show_velocity = show_velocity
wide.show_epoch = True
wide.show_framenumber = True
wide.show_lxly = True
wide.show_fxfy = True
renderlist = []
zoom_ts = zoom.fmf.get_all_timestamps().tolist()
df = madplot.load_bagfile_single_dataframe(rosbagf, arena, ffill=True,
tzname=TZNAME)
t0 = df.index[0].asm8.astype(np.int64) / 1e9
# find start and stop frames based on laser transition ----
laser_power = df['laser_power'].values
framenumber = df['h_framenumber'].values
laser_power = np.array( laser_power, copy=True )
laser_power[ np.isnan(laser_power) ] = 0
dlaser = laser_power[1:]-laser_power[:-1]
laser_transitions = np.nonzero(dlaser)[0]
startframenumber = -np.inf
stopframenumber = np.inf
if pre_frames is None:
assert post_frames is None
if pre_frames is not None:
assert post_frames is not None
for i,idx in enumerate(laser_transitions):
if i==0:
startframenumber = framenumber[idx-pre_frames]
elif i==1:
stopframenumber = framenumber[idx+post_frames]
else:
print 'ERROR: unknown laser transition in %r'%zoomf
moviemaker.cleanup()
return
# got start and stop frames -----
for idx,group in df.groupby('h_framenumber'):
# limit movies to only frames we want
this_framenumber = group['h_framenumber'].values[0]
if not (startframenumber <= this_framenumber and this_framenumber <= stopframenumber):
continue
#did we save a frame ?
try:
frameoffset = zoom_ts.index(idx)
except ValueError:
#missing frame (probbably because the video was not recorded at
#full frame rate
continue
frame = madplot.FMFFrame(offset=frameoffset, timestamp=idx)
row = group.dropna(subset=['tobj_id']).tail(1)
if len(row):
if maxt > 0:
dt = (row.index[0].asm8.astype(np.int64) / 1e9) - t0
if dt > maxt:
break
desc = madplot.FrameDescriptor(
None,
frame,
row,
row.index[0].asm8.astype(np.int64) / 1e9)
renderlist.append(desc)
if len(renderlist)==0:
moviemaker.cleanup()
return
wide.t0 = t0
panels = {}
#left half of screen
panels["wide"] = wide.get_benu_panel(
device_x0=0, device_x1=0.5*TARGET_OUT_W,
device_y0=0, device_y1=TARGET_OUT_H
)
panels["zoom"] = zoom.get_benu_panel(
device_x0=0.5*TARGET_OUT_W, device_x1=TARGET_OUT_W,
device_y0=0, device_y1=TARGET_OUT_H
)
actual_w, actual_h = benu.utils.negotiate_panel_size_same_height(panels, TARGET_OUT_W)
ass = Assembler(actual_w, actual_h,
panels,
wide,zoom,
moviemaker,
)
if not USE_MULTIPROCESSING:
pbar = madplot.get_progress_bar(moviemaker.movie_fname, len(renderlist))
for i,desc in enumerate(renderlist):
ass.render_frame(desc)
if not USE_MULTIPROCESSING:
pbar.update(i)
if not USE_MULTIPROCESSING:
pbar.finish()
if not os.path.exists(mp4_dir):
os.makedirs(mp4_dir)
moviefname = moviemaker.render(mp4_dir)
print "wrote", moviefname
moviemaker.cleanup()
def prepare_data(arena, path):
pool_df = madplot.load_bagfile_single_dataframe(path, arena, ffill=False, filter_short_pct=10.0)
targets = pool_df['target_type'].dropna().unique()
if len(targets) != 1:
raise Exception("Only head or body may be targeted")
if targets[0] == 1:
target_name = "head"
elif targets[0] == 2:
target_name = "body"
else:
raise Exception("Only head or body may be targeted")
ldf_all = pool_df[~pool_df['lobj_id'].isnull()]
hdf_all = pool_df[~pool_df['h_framenumber'].isnull()]
#the first fly targeted
prev = pool_df['lobj_id'].dropna().head(1)
prev_id = prev.values[0]
prev_ix = prev.index[0]
target_ranges = {
(0,100):[],
}
for ix, row in pool_df.iterrows():
lobj_id = row['lobj_id']
if np.isnan(lobj_id):
continue
elif lobj_id != prev_id:
df = pool_df[prev_ix:ix]
#mean speed of fly over its whole turn
v = df['v'].mean()
#details of the TrackedObj message
ldf = df[~df['lobj_id'].isnull()]
#we want to know when we go to TTM mode
gbs = ldf.groupby('mode')
try:
#this is the time we first switch to TTM
to_ttm = gbs.groups[2.0][0]
print "------------",v
ttm_errs = []
ttm_dts_thisfly = []
ttm_dts = []
wf_errs = []
wf_dts_thisfly = []
wf_dts = []
#get the head detection data once we start ttm
hdf = pool_df[to_ttm:ix]
#extract only the head messages
for i,(hix,hrow) in enumerate(hdf[~hdf['h_framenumber'].isnull()].iterrows()):
#skip when we miss the head
if hrow['head_x'] == 1e6:
continue
if target_name == "head":
ttm_err = math.sqrt( (hrow['head_x']-hrow['target_x'])**2 +\
(hrow['head_y']-hrow['target_y'])**2)
elif target_name == "body":
ttm_err = math.sqrt( (hrow['body_x']-hrow['target_x'])**2 +\
(hrow['body_y']-hrow['target_y'])**2)
#if the first value is < 10 it is likely actually associated
#with the last fly TTM, so ignore it....
if (i == 0) and (ttm_err < 100):
continue
#the time since we switched to this fly
thisfly_dt = (hix-prev_ix).total_seconds()
#the time since we switched to TTM
ttm_dt = (hix-to_ttm).total_seconds()
ttm_errs.append(ttm_err)
ttm_dts_thisfly.append(thisfly_dt)
ttm_dts.append(ttm_dt)
wdf = ldf[:hix].tail(1)
wf_err = math.sqrt( (wdf['fly_x']-wdf['laser_x'])**2 +\
(wdf['fly_y']-wdf['laser_y'])**2)
#the time since we switched to this fly
thisfly_dt = (wdf.index[0]-prev_ix).total_seconds()
#the time since we switched to TTM
ttm_dt = (wdf.index[0]-to_ttm).total_seconds()
wf_errs.append(wf_err)
wf_dts_thisfly.append(thisfly_dt)
wf_dts.append(ttm_dt)
print thisfly_dt, ttm_dt, ttm_err
ttm_err = Err(ttm_errs, ttm_dts_thisfly, ttm_dts)
wf_err = Err(wf_errs, wf_dts_thisfly, wf_dts)
trg = Target(lobj_id, prev_ix, ix, v, ttm_err, wf_err)
except KeyError:
#never switched to TTM
print "never switched to ttm"
trg = None
pass
#except Exception:
# print "UNKNOWN ERROR"
# trg = None
#classify the target into which speed range
if trg is not None:
for k in target_ranges:
vmin,vmax = k
if vmin < trg.v <= vmax:
target_ranges[k].append(trg)
prev_id = lobj_id
prev_ix = ix
data = {'target_name':target_name,
'pooldf':pool_df,
'ldf':ldf_all,
'hdf':hdf_all,
'target_ranges':target_ranges}
pkl_fname = "%s_%s.pkl" % (path, target_name)
cPickle.dump(data, open(pkl_fname,'wb'), -1)
return data
def doit_using_framenumber(user_data):
match, mdir, show_theta, show_velocity = user_data
zoomf = match.fmf
rosbagf = match.bag
maxt = match.maxt
arena = madplot.Arena(False)
zoom = madplot.FMFImagePlotter(zoomf, 'z_frame')
zoom.enable_color_correction(brightness=15, contrast=1.5)
wide = madplot.ArenaPlotter(arena)
wide.show_theta = show_theta
wide.show_velocity = show_velocity
wide.show_epoch = True
wide.show_framenumber = True
wide.show_lxly = True
wide.show_fxfy = True
renderlist = []
zoom_ts = zoom.fmf.get_all_timestamps().tolist()
df = madplot.load_bagfile_single_dataframe(rosbagf, arena, ffill=True)
t0 = df.index[0].asm8.astype(np.int64) / 1e9
for idx,group in df.groupby('h_framenumber'):
#did we save a frame ?
try:
frameoffset = zoom_ts.index(idx)
except ValueError:
#missing frame (probbably because the video was not recorded at
#full frame rate
continue
frame = madplot.FMFFrame(offset=frameoffset, timestamp=idx)
row = group.dropna(subset=['tobj_id']).tail(1)
if len(row):
if maxt > 0:
dt = (row.index[0].asm8.astype(np.int64) / 1e9) - t0
if dt > maxt:
break
desc = madplot.FrameDescriptor(
None,
frame,
row,
row.index[0].asm8.astype(np.int64) / 1e9)
renderlist.append(desc)
wide.t0 = t0
panels = {}
#left half of screen
panels["wide"] = wide.get_benu_panel(
device_x0=0, device_x1=0.5*TARGET_OUT_W,
device_y0=0, device_y1=TARGET_OUT_H
)
panels["zoom"] = zoom.get_benu_panel(
device_x0=0.5*TARGET_OUT_W, device_x1=TARGET_OUT_W,
device_y0=0, device_y1=TARGET_OUT_H
)
actual_w, actual_h = benu.utils.negotiate_panel_size_same_height(panels, TARGET_OUT_W)
moviemaker = madplot.MovieMaker(basename=os.path.basename(zoomf), fps=30)
target_moviefname = moviemaker.get_target_movie_name(mdir)
if os.path.exists(target_moviefname):
print 'target %r exists: skipping movie'%(target_moviefname,)
return
ass = Assembler(actual_w, actual_h,
panels,
wide,zoom,
moviemaker,
)
if not USE_MULTIPROCESSING:
pbar = madplot.get_progress_bar(moviemaker.movie_fname, len(renderlist))
for i,desc in enumerate(renderlist):
ass.render_frame(desc)
if not USE_MULTIPROCESSING:
pbar.update(i)
if not USE_MULTIPROCESSING:
pbar.finish()
if not os.path.exists(mdir):
os.makedirs(mdir)
moviefname = moviemaker.render(mdir)
print "wrote", moviefname
moviemaker.cleanup()
def doit_using_framenumber(user_data):
match, mp4_dir, show_theta, show_velocity, pre_frames, post_frames = user_data
zoomf = match.fmf
rosbagf = match.bag
maxt = match.maxt
moviemaker = madplot.MovieMaker(basename=os.path.basename(zoomf), fps=15)
target_moviefname = moviemaker.get_target_movie_name(mp4_dir)
if os.path.exists(target_moviefname):
print 'target %r exists: skipping movie' % (target_moviefname, )
moviemaker.cleanup()
return
arena = madplot.Arena(False)
zoom = madplot.FMFImagePlotter(zoomf, 'z_frame')
zoom.enable_color_correction(brightness=15, contrast=1.5)
wide = madplot.ArenaPlotter(arena, tzname=TZNAME)
wide.show_theta = show_theta
wide.show_velocity = show_velocity
wide.show_epoch = True
wide.show_framenumber = True
wide.show_lxly = True
wide.show_fxfy = True
renderlist = []
zoom_ts = zoom.fmf.get_all_timestamps().tolist()
df = madplot.load_bagfile_single_dataframe(rosbagf,
arena,
ffill=True,
tzname=TZNAME)
t0 = df.index[0].asm8.astype(np.int64) / 1e9
# find start and stop frames based on laser transition ----
laser_power = df['laser_power'].values
framenumber = df['h_framenumber'].values
laser_power = np.array(laser_power, copy=True)
laser_power[np.isnan(laser_power)] = 0
dlaser = laser_power[1:] - laser_power[:-1]
laser_transitions = np.nonzero(dlaser)[0]
startframenumber = -np.inf
stopframenumber = np.inf
if pre_frames is None:
assert post_frames is None
if pre_frames is not None:
assert post_frames is not None
for i, idx in enumerate(laser_transitions):
if i == 0:
startframenumber = framenumber[idx - pre_frames]
elif i == 1:
stopframenumber = framenumber[idx + post_frames]
else:
print 'ERROR: unknown laser transition in %r' % zoomf
moviemaker.cleanup()
return
# got start and stop frames -----
for idx, group in df.groupby('h_framenumber'):
# limit movies to only frames we want
this_framenumber = group['h_framenumber'].values[0]
if not (startframenumber <= this_framenumber
and this_framenumber <= stopframenumber):
continue
#did we save a frame ?
try:
frameoffset = zoom_ts.index(idx)
except ValueError:
#missing frame (probbably because the video was not recorded at
#full frame rate
continue
frame = madplot.FMFFrame(offset=frameoffset, timestamp=idx)
row = group.dropna(subset=['tobj_id']).tail(1)
if len(row):
if maxt > 0:
dt = (row.index[0].asm8.astype(np.int64) / 1e9) - t0
if dt > maxt:
break
desc = madplot.FrameDescriptor(
None, frame, row, row.index[0].asm8.astype(np.int64) / 1e9)
renderlist.append(desc)
if len(renderlist) == 0:
moviemaker.cleanup()
return
wide.t0 = t0
panels = {}
#left half of screen
panels["wide"] = wide.get_benu_panel(device_x0=0,
device_x1=0.5 * TARGET_OUT_W,
device_y0=0,
device_y1=TARGET_OUT_H)
panels["zoom"] = zoom.get_benu_panel(device_x0=0.5 * TARGET_OUT_W,
device_x1=TARGET_OUT_W,
device_y0=0,
device_y1=TARGET_OUT_H)
actual_w, actual_h = benu.utils.negotiate_panel_size_same_height(
panels, TARGET_OUT_W)
ass = Assembler(
actual_w,
actual_h,
panels,
wide,
zoom,
moviemaker,
)
if not USE_MULTIPROCESSING:
pbar = madplot.get_progress_bar(moviemaker.movie_fname,
len(renderlist))
for i, desc in enumerate(renderlist):
ass.render_frame(desc)
if not USE_MULTIPROCESSING:
pbar.update(i)
if not USE_MULTIPROCESSING:
pbar.finish()
if not os.path.exists(mp4_dir):
os.makedirs(mp4_dir)
moviefname = moviemaker.render(mp4_dir)
print "wrote", moviefname
moviemaker.cleanup()
def prepare_data(arena, path, smoothstr, smooth):
path = os.path.abspath(path)
print 'opening bag files in %r'%path
GENOTYPES = {"NINE":"*.bag",
"CSSHITS":"ctrl/CSSHITS/*.bag",
"NINGal4":"ctrl/NINGal4/*.bag"}
CHUNKS = collections.OrderedDict()
CHUNKS["1_loff+ve"] = (0,30)
CHUNKS["2_loff-ve"] = (30,60)
CHUNKS["3_lon-ve"] = (60,120)
CHUNKS["4_lon+ve"] = (120,180)
CHUNKS["5_loff+ve"] = (180,240)
CHUNKS["6_loff-ve"] = (240,300)
CHUNKS["7_loff+ve"] = (300,360)
save_times = None
data = {gt:dict(chunk={}) for gt in GENOTYPES}
for gti,gt in enumerate(GENOTYPES):
#if gti>=1: break
pattern = os.path.join(path, GENOTYPES[gt])
bags = glob.glob(pattern)
for bag in bags:
df = madplot.load_bagfile_single_dataframe(
bag, arena,
ffill=['laser_power','e_rotator_velocity_data'],
extra_topics={'/rotator/velocity':['data']},
smooth=smooth
)
supplement_angles(df)
supplement_times(df)
if 1:
# John's category stuff
#there might be multiple object_ids in the same bagfile for the same
#fly, so calculate dtheta for all of them
dts = []
for obj_id, group in df.groupby('tobj_id'):
if np.isnan(obj_id):
continue
dts.append( calc_dtheta(group) )
#join the dthetas vertically (axis=0) and insert as a column into
#the full dataframe
df['dtheta'] = pd.concat(dts, axis=0)
good = ~np.isnan(df['time_since_start'])
if save_times is None:
save_times = np.arange(0,
(df['time_since_start'][good].values[-1]+1e-6),
1.0 ) # every second
if 'timeseries_angular_vel' not in data[gt]:
data[gt]['bag_fname'] = []
data[gt]['timeseries_angular_vel'] = []
data[gt]['timeseries_vel'] = []
data[gt]['save_times'] = save_times
good_stim = ~np.isnan(df['e_rotator_velocity_data'].values)
good_both = good & good_stim
data[gt]['stimulus_velocity'] = (
df['time_since_start'].values[good_both],
df['e_rotator_velocity_data'].values[good_both])
if gt=='CSSHITS' and 'laser_power' not in data[gt]:
interp = scipy.interpolate.interp1d( df['time_since_start'][good],
df['laser_power'][good],
bounds_error=False,
fill_value=np.nan,
)
save_laser = interp(save_times)
data[gt]['laser_power'] = save_laser
interp = scipy.interpolate.interp1d( df['time_since_start'][good],
df['angular_velocity'][good],
bounds_error=False,
fill_value=np.nan,
)
save_angular_vel = interp(save_times)
data[gt]['timeseries_angular_vel'].append( save_angular_vel )
data[gt]['bag_fname'].append( bag )
interp = scipy.interpolate.interp1d( df['time_since_start'][good],
df['v'][good],
bounds_error=False,
fill_value=np.nan,
)
save_vel = interp(save_times)
data[gt]['timeseries_vel'].append( save_vel )
# if 1:
# break
#calculate mean_dtheta over each phase of the experiment
t00 = df.index[0]
for c in CHUNKS:
s0,s1 = CHUNKS[c]
t0 = datetime.timedelta(seconds=s0)
t1 = datetime.timedelta(seconds=s1)
mean_dtheta = df[t00+t0:t00+t1]["dtheta"].mean()
mean_v = df[t00+t0:t00+t1]["v"].mean()
if c not in data[gt]:
data[gt]['chunk'][c] = []
data[gt]['chunk'][c].append( (mean_dtheta, mean_v) )
pickle.dump(data, open(CACHE_FNAME,'wb'), -1)
print 'saved cache to %s'%CACHE_FNAME
return data
def doit_using_framenumber(user_data):
match, mdir, show_theta, show_velocity = user_data
zoomf = match.fmf
rosbagf = match.bag
maxt = match.maxt
arena = madplot.Arena(False)
zoom = madplot.FMFImagePlotter(zoomf, 'z_frame')
zoom.enable_color_correction(brightness=15, contrast=1.5)
wide = madplot.ArenaPlotter(arena)
wide.show_theta = show_theta
wide.show_velocity = show_velocity
wide.show_epoch = True
wide.show_framenumber = True
wide.show_lxly = True
wide.show_fxfy = True
renderlist = []
zoom_ts = zoom.fmf.get_all_timestamps().tolist()
df = madplot.load_bagfile_single_dataframe(rosbagf, arena, ffill=True)
t0 = df.index[0].asm8.astype(np.int64) / 1e9
for idx, group in df.groupby('h_framenumber'):
#did we save a frame ?
try:
frameoffset = zoom_ts.index(idx)
except ValueError:
#missing frame (probbably because the video was not recorded at
#full frame rate
continue
frame = madplot.FMFFrame(offset=frameoffset, timestamp=idx)
row = group.dropna(subset=['tobj_id']).tail(1)
if len(row):
if maxt > 0:
dt = (row.index[0].asm8.astype(np.int64) / 1e9) - t0
if dt > maxt:
break
desc = madplot.FrameDescriptor(
None, frame, row, row.index[0].asm8.astype(np.int64) / 1e9)
renderlist.append(desc)
wide.t0 = t0
panels = {}
#left half of screen
panels["wide"] = wide.get_benu_panel(device_x0=0,
device_x1=0.5 * TARGET_OUT_W,
device_y0=0,
device_y1=TARGET_OUT_H)
panels["zoom"] = zoom.get_benu_panel(device_x0=0.5 * TARGET_OUT_W,
device_x1=TARGET_OUT_W,
device_y0=0,
device_y1=TARGET_OUT_H)
actual_w, actual_h = benu.utils.negotiate_panel_size_same_height(
panels, TARGET_OUT_W)
moviemaker = madplot.MovieMaker(basename=os.path.basename(zoomf), fps=30)
target_moviefname = moviemaker.get_target_movie_name(mdir)
if os.path.exists(target_moviefname):
print 'target %r exists: skipping movie' % (target_moviefname, )
return
ass = Assembler(
actual_w,
actual_h,
panels,
wide,
zoom,
moviemaker,
)
if not USE_MULTIPROCESSING:
pbar = madplot.get_progress_bar(moviemaker.movie_fname,
len(renderlist))
for i, desc in enumerate(renderlist):
ass.render_frame(desc)
if not USE_MULTIPROCESSING:
pbar.update(i)
if not USE_MULTIPROCESSING:
pbar.finish()
if not os.path.exists(mdir):
os.makedirs(mdir)
moviefname = moviemaker.render(mdir)
print "wrote", moviefname
moviemaker.cleanup()
parser.add_argument('--outdir', help='destination directory for mp4')
args = parser.parse_args()
BAG_FILE = args.path[0]
ZOOM_FMF = args.zoom_fmf
WIDE_FMF = args.wide_fmf
wfmf = madplot.FMFTrajectoryPlotter(WIDE_FMF)
zfmf = madplot.FMFTTLPlotter(ZOOM_FMF)
zfmf.enable_color_correction(brightness=20, contrast=1.5)
arena = madplot.Arena(False)
print "loading data"
df = madplot.load_bagfile_single_dataframe(BAG_FILE, arena, ffill=True)
wt = wfmf.fmf.get_all_timestamps()
zt = zfmf.fmf.get_all_timestamps()
if len(df['tobj_id'].dropna().unique()) != 1:
print "TTM movies require single unique object IDs, I think..."
sys.exit(1)
frames = build_framedesc_list(df, wt, zt)
if not frames:
print "no frames to render"
sys.exit(0)
else:
print len(frames),"frames to render"
moviet0 = frames[0].epoch
