offline_tracking_sample.py

from ethoscope.hardware.input.cameras import *

from ethoscope.trackers.adaptive_bg_tracker import AdaptiveBGModel
from ethoscope.trackers.multi_fly_tracker import MultiFlyTracker, HaarTracker
from ethoscope.drawers.drawers import DefaultDrawer

from ethoscope.core.monitor import Monitor
from ethoscope.roi_builders.roi_builders import  DefaultROIBuilder

from ethoscope.utils.io import rawdatawriter, npyAppendableFile


# the input video - we work on an mp4 acquired with the Record function
#input_video = "/home/gg/Downloads/test_video.mp4"

l = []; input_video_list = open("/home/gg/OLFACTOVIDEOS/file_list.txt")
for input_video_file in input_video_list: l.append(input_video_file)

for input_video in l[6:]:
    input_video = input_video.strip()

    #the location of the haar cascade file (in case we want to use the haar tracking function)
    cascade = "/home/gg/Cabinet/Work/Projects/custom_devices/ethoscope/olfactometer/olfactometer_tracking/data/cascade.xml"

    #how many flies we expect
    entities=42

    #the type of tracking we want to use
    ttype = "haar"

    #Choice of trackers
    # default - AdaptiveBGModel - The default tracker for fruit flies. One animal per ROI.
    # multi - MultiFlyTracker - An experimental tracker to monitor several animals per ROI.
    # haar - HaarTracker - An experimental tracker to monitor several animals per ROI using a Haar Cascade.

    # the output video - we can save a video with the drawing of the tracking
    output_video = None
    #output_video = "/home/gg/tracked_video_big_fly.mp4"

    camera = MovieVirtualCamera(input_video)

    # we use the default drawer and we show the video as we track - this is useful to understand how things are going
    # disabling the video will speed things up. We can also save a video output if we pass a filename for the video
    drawer = None
    #drawer = DefaultDrawer(draw_frames = True, video_out = output_video, video_out_fps=25)


    #Choice of ROIs
    roi_builder = DefaultROIBuilder() # One Big ROI using the Default ROIBuilder
    #roi_builder = SleepMonitorWithTargetROIBuilder() # the default ROI structure for a 20 vials arena. Requires targets
    #roi_builder = ImgMaskROIBuilder("maskfile.png") # Creates a custom MASK as ROI taken from a greyscale image file. Each different shade of grey is a different ROI


    #create the rois
    rois = roi_builder.build(camera)

    # We use the npy tracker to save data in a npy file
    rdw = rawdatawriter(basename = input_video + ".npy", n_rois=len(rois), entities=entities)
    #rdw = None


    #for multifly tracking using BS subtraction

    if ttype == "multi":
        monit = Monitor(camera, MultiFlyTracker, rois, stimulators=None, data={ 'maxN' : 50, 
                                            'visualise' : False ,
                                            'fg_data' : { 'sample_size' : 400, 'normal_limits' : (50, 200), 'tolerance' : 0.8 }
                                          } )

    #For the haar tracking
    if ttype == "haar":
        monit = Monitor(camera, HaarTracker, rois, stimulators=None, data = { 'maxN' : entities, 
                         'cascade' : cascade,
                         'scaleFactor' : 1.1,
                         'minNeighbors' : 3,
                         'flags' : 0,
                         'minSize' : (15,15),
                         'maxSize' : (20,20),
                         'visualise' : False }
                         )

    if ttype == "default":
        monit = Monitor(camera, AdaptiveBGModel, rois)

    # Starts the tracking monitor
    monit.run(drawer = drawer, result_writer = rdw, verbose = True)

    #convert the anpy to npy
    print ("Tracking done, now converting")
    fh = npyAppendableFile (input_video + "_001.anpy")
    fh.convert()