n the training process, the center loss becomes larger and larger. What is the mistake

[alanMachineLeraning]

[fisakhan]

My be your training dataset is too small.

[jjsjunior]

Are you shuffling your trainset? I had this problem and after I starting shuffling the issue was resolved.

[RyanCV]

@jjsjunior do you mean shuffle here: index_queue = tf.train.range_input_producer(range_size, num_epochs=None, shuffle=True, seed=None, capacity=32), I have tried "True" and "False", but it didn't help.

@fisakhan : what do you mean by the training dataset is too small? I have 11k images (374 classes) for training, but I found the value in some part of the centers matrix (dimension is 374x128) become very large, any reason? Thanks.

[RyanCV]

@alanMachineLeraning How did you solve this problem of using center loss during training? I also have this kind of inf of center loss after 3 epochs during finetuning of facenet.

[jetsmith]

@RyanCV @alanMachineLeraning i have the same problem, have you solved it? How?