In [26]:

    

# import multi tracker analysis


    

In [27]:

    

# load hdf5 data as a pandas dataframe, and load the configuration
# change the argument to match your data path!
pd, config = mta.read_hdf5_file_to_pandas.load_and_preprocess_data('/home/lab/demo/demo_1/retracked/20170711_150010_N1/20170717_105053_N1_trackedobjects.hdf5')


    

    




Preprocessing data - see config file for details!
[]
Found too many, or too few files
No delete cut join instructions found in path!

In [28]:

    

# pandas dataframe structure: each row is a measurement corresponding to one of the objects for each frame
pd.keys()


    

    
Out[28]:





Index([u'angle', u'area', u'frames', u'measurement_x', u'measurement_y',
       u'objid', u'position_x', u'position_y', u'time_epoch_nsecs',
       u'time_epoch_secs', u'velocity_x', u'velocity_y', u'time_epoch',
       u'speed'],
      dtype='object')

In [29]:

    

# show the object id's
import numpy as np
np.unique(pd.objid)


    

    
Out[29]:





array([    0,     1,     2, ..., 53569, 53570, 53571])

In [30]:

    

position_for_10 = pd[pd.objid==10].position_x.values
position_for_10


    

    
Out[30]:





array([ 1424.,  1424.,  1424.])

In [31]:

    

# wrap pandas dataframe into a dataset object for intuitive processing
# note: using copy=True creates a copy of the data
#       this allows you to add attributes to trajecties and pickle the dataset structure for future use
dataset = mta.read_hdf5_file_to_pandas.Dataset(pd, config=config)
dataset.load_keys() # load the object ids
trajec = dataset.trajec(dataset.keys[7]) # load the 7th trajectory
trajec.position_x # show the position x data


    

    
Out[31]:





array([ 612.        ,  612.        ,  612.        ,  620.04771372])

In [32]:

    

import matplotlib.pyplot as plt
%matplotlib inline


    

In [33]:

    

# plot the trajectory; this requires downloading my plotting library: https://github.com/florisvb/FlyPlotLib
# make a heatmap of the trajectories
bgimg = mta.plot.get_filename(config.path, 'bgimg')
binsx, binsy = mta.plot.get_bins_from_backgroundimage(bgimg)
mta.plot.plot_trajectories(pd, binsx, binsy)


    

In [34]:

    

mta.plot.plot_heatmap(pd, binsx, binsy)


    

    
Out[34]:





<matplotlib.axes._subplots.AxesSubplot at 0x7ff8e2537b50>






    

In [ ]:

    

#probability of crossing to lf or rt v. time
#each objid doesn't have a recorded positon for entire duration of video 
#access time and align array? add 0s in the gaps
rt_crossed = []
lf_crossed = []
for i in range(len(np.unique(pd.objid))):
        trajec = dataset.trajec(dataset.keys[i])
        rt_crossing=[]
        lf_crossing=[]
        for i in range(len(trajec.position_x)):
                cross = []
                if trajec.position_x[i]<700:
                        cross.extend(0)
                elif trajec.position_x[i]>700:
                        cross.extend(1)
                        for i in range(len(cross))
                                if cross[i+1]-cross[i]=0
                                        rt_crossing.extend(0)
                                        lf_crossing.extend(0)
                                if cross[i+1]-cross[i]=1
                                        rt_crossing.extend(1)
                                        lt_crossing.extend(0)
                                elif cross[i+1]-cross[i]=-1
                                        rt_crossing.extend(0)
                                        lf_crossing.extend(1)

        rt_crossed.append[rt_crossing]
        lf_crossed.append[lf_crossing]
prob_rt=[]
prob_lf=[]
for i in range(len(cross)):
        for ii in range(len(cross)):
                sum_rt = 0
                sum_lf = 0
                sum_rt = sum_rt + rt_crossed[ii][i]
                sum_lf = sum_lf + lf_crossed[ii][i]
                prob_cross_rt = 0
                prob_cross_lf = 0
                prob_cross_rt = sum_rt/len(crossed)
                prob_cross_lf = -1* sum_lf/len(crossed)
                prob_rt.extend(prob_cross_rt)
                prob_lf.extend(prob_cross_lf)
plt.plot(prob_rt)
plt.plot(prob_lf)
plt.ylabel('probability of crossing (+)right (-)left')
plt.xlabel('time')
odor_pulse = []
lf_edge=0
rt_edge=2
for i in range(int(math.ceil(frames/30))):
        lf_edge= lf_edge+28
        rt_edge= rt_edge+28
        plt.axvspan(lf_edge, rt_edge, 0, 1, alpha=0.5, color= 'yellow')
plt.show()


    

In [ ]:

    

#lf & rt occupancy v. time
#same problem as above
b = []
position = []
trajec = dataset.trajec(dataset.keys[5])
time = len(trajec.position_x)
for i in range(len(np.unique(pd.objid))):
        trajec = dataset.trajec(dataset.keys[i])
        position.append(trajec.position_x)
for i in range (time):
        rt = (np.sum(position[:,i]<700))
        lf = len(position)-rt
        b.append([lf,rt])
        rt =0
        lf =0
#plot blobs on lf or rt w/ odor pulses
plt.plot(b)
plt.ylabel('flies')
plt.xlabel('time')
odor_pulse = []
lf_edge=0
rt_edge=2
#y_max = max(b)  
#print y_max
for i in range(int(math.ceil(frames/30))):
        lf_edge= lf_edge+28
        rt_edge= rt_edge+28
        plt.axvspan(lf_edge, rt_edge, 0, 40, alpha=0.5, color= 'yellow')
plt.show()


    

In [ ]:

    

#avg walking speed v. time
#use .speed?
#assuming that once an object get tracked, it stays until end of video
speed = []
max_len = 0
for i in range (len(np.unique(pd.objid))):
    trajec = dataset.trajec(dataset.keys[i])
    speed.append(trajec.speed)
for i in range (len(speed)):
    curr_len  = len(speed[i ,:])
    if curr_len > max_len:
        max_len = curr_len