Function lanenet_model::lanenet_discriminative_loss::discriminative_loss_single

• Defined in File lanenet_discriminative_loss.py

Function Documentation

lanenet_model.lanenet_discriminative_loss.discriminative_loss_single(prediction, correct_label, feature_dim, label_shape, delta_v, delta_d, param_var, param_dist, param_reg)

Instance segmentation loss function mentioned in equ(1)

Parameters

• prediction – inference of network

• correct_label – instance label

• feature_dim – feature dimension of prediction

• label_shape – shape of label

• delta_v – cut off variance distance

• delta_d – cut off cluster distance

• param_var – weight for intra cluster variance

• param_dist – weight for inter cluster distances

• param_reg – weight regularization