DCD: Deep Crossing detector

Given a collection of Blob objects (see Blob), the crossing detector module allows to apply a pre-computed model of the area to each of the blobs and, if specified, use a function approximator (in this case a convolutional neural network) in order to distinguish between Blob representing single individual and touching individuals.

crossing_detector.detect_crossings(list_of_blobs, video, model_area, use_network=True, return_store_objects=False, plot_flag=True)[source]

Short summary.

list_of_blobs : <ListOfBlobs object>

Collection of the Blob objects extracted from the video

video : <Video object>

Object containing all the parameters of the video.

model_area : function

Model of the area of a single individual

use_network : bool

If True the Deep Crossing Detector is used to distinguish between individuals and crossings images. Otherwise only the model area is applied

return_store_objects : bool

If True the instantiations of the class Store_Accuracy_and_Loss are returned by the function

plot_flag : bool

If True a figure representing the values of the loss function, accuracy and accuracy per class for both the training and validation set.