Upload 2 files
Browse files- README.md +71 -0
- arcface.onnx +3 -0
README.md
ADDED
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
library_name: diffusers
|
6 |
+
---
|
7 |
+
|
8 |
+
# Arc2Face Model Card
|
9 |
+
|
10 |
+
<div align="center">
|
11 |
+
|
12 |
+
[**Project Page**](https://arc2face.github.io/) **|** [**Paper (ArXiv)**](https://arxiv.org/abs/2403.11641) **|** [**Code**](https://github.com/foivospar/Arc2Face) **|** [🤗 **Gradio demo**](https://huggingface.co/spaces/FoivosPar/Arc2Face)
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
</div>
|
17 |
+
|
18 |
+
## Introduction
|
19 |
+
|
20 |
+
Arc2Face is an ID-conditioned face model, that can generate diverse, ID-consistent photos of a person given only its ArcFace ID-embedding.
|
21 |
+
It is trained on a restored version of the WebFace42M face recognition database, and is further fine-tuned on FFHQ and CelebA-HQ.
|
22 |
+
|
23 |
+
<div align="center">
|
24 |
+
<img src='assets/samples_short.jpg'>
|
25 |
+
</div>
|
26 |
+
|
27 |
+
## Model Details
|
28 |
+
|
29 |
+
It consists of 2 components:
|
30 |
+
- encoder, a finetuned CLIP ViT-L/14 model
|
31 |
+
- arc2face, a finetuned UNet model
|
32 |
+
|
33 |
+
both of which are fine-tuned from [runwayml/stable-diffusion-v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5).
|
34 |
+
The encoder is tailored for projecting ID-embeddings to the CLIP latent space.
|
35 |
+
Arc2Face adapts the pre-trained backbone to the task of ID-to-face generation, conditioned solely on ID vectors.
|
36 |
+
|
37 |
+
## Usage
|
38 |
+
|
39 |
+
The models can be downloaded directly from this repository or using python:
|
40 |
+
```python
|
41 |
+
from huggingface_hub import hf_hub_download
|
42 |
+
|
43 |
+
hf_hub_download(repo_id="FoivosPar/Arc2Face", filename="arc2face/config.json", local_dir="./models")
|
44 |
+
hf_hub_download(repo_id="FoivosPar/Arc2Face", filename="arc2face/diffusion_pytorch_model.safetensors", local_dir="./models")
|
45 |
+
hf_hub_download(repo_id="FoivosPar/Arc2Face", filename="encoder/config.json", local_dir="./models")
|
46 |
+
hf_hub_download(repo_id="FoivosPar/Arc2Face", filename="encoder/pytorch_model.bin", local_dir="./models")
|
47 |
+
```
|
48 |
+
|
49 |
+
Please check our [GitHub repository](https://github.com/foivospar/Arc2Face) for complete inference instructions.
|
50 |
+
|
51 |
+
## Limitations and Bias
|
52 |
+
|
53 |
+
- Only one person per image can be generated.
|
54 |
+
- Poses are constrained to the frontal hemisphere, similar to FFHQ images.
|
55 |
+
- The model may reflect the biases of the training data or the ID encoder.
|
56 |
+
|
57 |
+
## Citation
|
58 |
+
|
59 |
+
|
60 |
+
**BibTeX:**
|
61 |
+
|
62 |
+
```bibtex
|
63 |
+
@misc{paraperas2024arc2face,
|
64 |
+
title={Arc2Face: A Foundation Model of Human Faces},
|
65 |
+
author={Foivos Paraperas Papantoniou and Alexandros Lattas and Stylianos Moschoglou and Jiankang Deng and Bernhard Kainz and Stefanos Zafeiriou},
|
66 |
+
year={2024},
|
67 |
+
eprint={2403.11641},
|
68 |
+
archivePrefix={arXiv},
|
69 |
+
primaryClass={cs.CV}
|
70 |
+
}
|
71 |
+
```
|
arcface.onnx
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ec639a0429b4819130d1405a2d3b38beaa4cc4a6c5bd9cf48b94fdf65461de83
|
3 |
+
size 260694151
|