search

sinhasam / vaal

Variational Adversarial Active Learning (ICCV 2019)
https://arxiv.org/abs/1904.00370
BSD 2-Clause "Simplified" License
225 stars 51 forks source link

Network Architecture #22

Open spoorgholi74 opened 4 years ago

spoorgholi74 commented 4 years ago

Hi, I was wondering why did you use a VAE to get the embedding instead of just the output of a middle layer from a CNN?

SaynaEbrahimi commented 4 years ago

Because VAE would be trained with a reconstruction loss which not only ensures input information are preserved but also creates a rich low dimensional latent space on which the discriminator would be trained better.