Latent space extraction of compression models
68 views
Skip to first unread message
lorenzo sibi
Oct 16, 2023, 9:09:58 AM10/16/23
to tensorflow-compression
Hello team tf-compression,
I am conducting a magisterial thesis regarding the study of the effects of AI-based compression methods on deepfake detention. To continue, I am looking for a way to get the latent space representation of each individual model (HiFiC, ms2020-cc10-mse-[1-10], etc...) for each compressed image used in my research.
I noticed the presence of the "dump" command for the tfci.py script, but the name of the tensor to be extracted must be specified, but the list is huge plus the architecture and implementation of each individual model should be known.
Could you give me a suggestion or a list of possible names (I hope I am not asking too much :) ) of the required tensors?
Thank you very much in advance, I take this opportunity to thank the team for doing such splendid work. :)
I am conducting a magisterial thesis regarding the study of the effects of AI-based compression methods on deepfake detention. To continue, I am looking for a way to get the latent space representation of each individual model (HiFiC, ms2020-cc10-mse-[1-10], etc...) for each compressed image used in my research.
I noticed the presence of the "dump" command for the tfci.py script, but the name of the tensor to be extracted must be specified, but the list is huge plus the architecture and implementation of each individual model should be known.
Could you give me a suggestion or a list of possible names (I hope I am not asking too much :) ) of the required tensors?
Thank you very much in advance, I take this opportunity to thank the team for doing such splendid work. :)
Reply all
Reply to author
Forward
0 new messages