Set up to replicate the reenactment results in the paper

shrubb / latent-pose-reenactment

The authors' implementation of the "Neural Head Reenactment with Latent Pose Descriptors" (CVPR 2020) paper.

https://shrubb.github.io/research/latent-pose-reenactment/

Apache License 2.0

180 stars 34 forks source link

Set up to replicate the reenactment results in the paper #7

Closed Blade6570 closed 3 years ago

Blade6570 commented 3 years ago

Hi, I am trying to replicate the pose reconstruction error and identity error reported in the paper. The instruction in this repo involves some amount of user's judgement in the fine tuning to get the best results.

Is there any default settings to generate the results in the paper? That would be good to test this model on number of videos for a fair comparison to other counter parts.

Thanks

shrubb commented 3 years ago

Hi! You can check the actual code that we used to compute metrics. It has those parameters. 'batched_drive.py' and 'compute_pose_identity_error.py' may also help.

Blade6570 commented 3 years ago

Thanks for pointing to those files. It is working perfectly.