I think the labels for classifier should be ones during information negative mining - Githubissues

nuclearboy95 / Anomaly-Detection-OCGAN-tensorflow

Tensorflow implementation of OCGAN: One-class Novelty Detection Using GANs with Constrained Latent Representations

22 stars 6 forks source link

I think the labels for classifier should be ones during information negative mining #6

Open mshakeelt opened 1 year ago

mshakeelt commented 1 year ago

https://github.com/nuclearboy95/Anomaly-Detection-OCGAN-tensorflow/blob/6be4602dacfe547077cd6ce2dd14e85cd3e84355/ocgan/ocgan.py#L62

mshakeelt commented 1 year ago

And also the loss should be calculated inside the loop, not only once!

mshakeelt commented 1 year ago

I am referring to the algorithm proposed in the original paper at page 6 (https://openaccess.thecvf.com/content_CVPR_2019/papers/Perera_OCGAN_One-Class_Novelty_Detection_Using_GANs_With_Constrained_Latent_Representations_CVPR_2019_paper.pdf)