Kaede Shiohara

Thank you for your interest in our work. This is because the face recognition (identity encoding) part of our code is based on [insightface](https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch) that uses 112x112 images. In practice,...

Exactly. 112x112 images are considered sufficient for face recognition, but not obvious for generative purposes. Higher resolution may yield better results for generative models as you mentioned.

Instead of building the docker image with `build.sh`, you can pull the built image directly from: https://hub.docker.com/r/mapooon/sbi.

Try to install packages with `requirements.txt` we released at: https://drive.google.com/file/d/14ZjiBFG825qP8ZKE_XGzQ6yob4XHx_Ml/view?usp=share_link After executing the docker, you can use it as follows: `pip install -r requirements.txt`. Please let me know if you...

Thanks to an enthusiastic collaborator, we have found that there is a bug in crop_dlib_ff.py. And we have just fixed the bug, so please try again from the point of...

Hi, The performance of the classifier strongly relies on the preprocess, e.g., cropping. We recommend you use the same preprocess as we provide in this repository.

It seems that the CelebDF videos were not loaded properly. Place them following `data/datasets.md`.

CelebDFv2 has 518 test videos. So I think there is no problem with the txt file. Please check if videos can be loaded by other ways. For example, by executing...

Sorry for the late reply. We don't release the docker image due to security problems now. Alternatively, we release `requirements.txt` [here](https://drive.google.com/file/d/14ZjiBFG825qP8ZKE_XGzQ6yob4XHx_Ml/view?usp=share_link). The docker version we use is `20.10.8`.

Hi, We provide efnb4 weights trained on SBIs. If you train other models(e.g., ResNet34) on SBIs, you need to - start with imagenet pretrained weights or other pretrained ones because...