Implementation of our paper 'PixelLink: Detecting Scene Text via Instance Segmentation' in AAAI2018
768
stars
255
forks
source link
怀疑regularization_loss = tf.add_n(tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)) #124
Closed
ligw0329 closed 5 years ago
没看到有正则损失,为什么会把这个regularization_loss代入总损失进行训练?