Skip to main content

Face related toolkit

Project description

FACER

Face related toolkit. This repo is still under construction to include more models.

Updates

  • [14/05/2023] Face attribute recognition model trained on CelebA is available, check it out here.
  • [04/05/2023] Face alignment model trained on IBUG300W, AFLW19, WFLW dataset is available, check it out here.
  • [27/04/2023] Face parsing model trained on CelebM dataset is available, check it out here.

Install

The easiest way to install it is using pip:

pip install git+https://github.com/FacePerceiver/facer.git@main

No extra setup needs, pretrained weights will be downloaded automatically.

If you have trouble install from source, you can try install from PyPI:

pip install pyfacer

the PyPI version is not guaranteed to be the latest version, but we will try to keep it up to date.

Face Detection

We simply wrap a retinaface detector for easy usage.

import facer

image = facer.hwc2bchw(facer.read_hwc('data/twogirls.jpg')).to(device=device)  # image: 1 x 3 x h x w

face_detector = facer.face_detector('retinaface/mobilenet', device=device)
with torch.inference_mode():
    faces = face_detector(image)

facer.show_bchw(facer.draw_bchw(image, faces))

Check this notebook for full example.

Please consider citing

@inproceedings{deng2020retinaface,
  title={Retinaface: Single-shot multi-level face localisation in the wild},
  author={Deng, Jiankang and Guo, Jia and Ververas, Evangelos and Kotsia, Irene and Zafeiriou, Stefanos},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={5203--5212},
  year={2020}
}

Face Parsing

We wrap the FaRL models for face parsing.

import torch
import facer

device = 'cuda' if torch.cuda.is_available() else 'cpu'

image = facer.hwc2bchw(facer.read_hwc('data/twogirls.jpg')).to(device=device)  # image: 1 x 3 x h x w

face_detector = facer.face_detector('retinaface/mobilenet', device=device)
with torch.inference_mode():
    faces = face_detector(image)

face_parser = facer.face_parser('farl/lapa/448', device=device) # optional "farl/celebm/448"

with torch.inference_mode():
    faces = face_parser(image, faces)

seg_logits = faces['seg']['logits']
seg_probs = seg_logits.softmax(dim=1)  # nfaces x nclasses x h x w
n_classes = seg_probs.size(1)
vis_seg_probs = seg_probs.argmax(dim=1).float()/n_classes*255
vis_img = vis_seg_probs.sum(0, keepdim=True)
facer.show_bhw(vis_img)
facer.show_bchw(facer.draw_bchw(image, faces))

Check this notebook for full example.

Please consider citing

@inproceedings{zheng2022farl,
  title={General facial representation learning in a visual-linguistic manner},
  author={Zheng, Yinglin and Yang, Hao and Zhang, Ting and Bao, Jianmin and Chen, Dongdong and Huang, Yangyu and Yuan, Lu and Chen, Dong and Zeng, Ming and Wen, Fang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={18697--18709},
  year={2022}
}

Face Alignment

We wrap the FaRL models for face alignment.

import torch
import cv2
from matplotlib import pyplot as plt

device = 'cuda' if torch.cuda.is_available() else 'cpu'

import facer
img_file = 'data/twogirls.jpg'
# image: 1 x 3 x h x w
image = facer.hwc2bchw(facer.read_hwc(img_file)).to(device=device)  

face_detector = facer.face_detector('retinaface/mobilenet', device=device)
with torch.inference_mode():
    faces = face_detector(image)

face_aligner = facer.face_aligner('farl/ibug300w/448', device=device) # optional: "farl/wflw/448", "farl/aflw19/448"

with torch.inference_mode():
    faces = face_aligner(image, faces)

img = cv2.imread(img_file)[..., ::-1]
vis_img = img.copy()
for pts in faces['alignment']:
    vis_img = facer.draw_landmarks(vis_img, None, pts.cpu().numpy())
plt.imshow(vis_img)

Check this notebook for full example.

Please consider citing

@inproceedings{zheng2022farl,
  title={General facial representation learning in a visual-linguistic manner},
  author={Zheng, Yinglin and Yang, Hao and Zhang, Ting and Bao, Jianmin and Chen, Dongdong and Huang, Yangyu and Yuan, Lu and Chen, Dong and Zeng, Ming and Wen, Fang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={18697--18709},
  year={2022}
}

Face Attribute Recognition

We wrap the FaRL models for face attribute recognition, the model achieves 92.06% accuracy on CelebA dataset.

import sys
import torch
import facer

device = "cuda" if torch.cuda.is_available() else "cpu"

# image: 1 x 3 x h x w
image = facer.hwc2bchw(facer.read_hwc("data/girl.jpg")).to(device=device)

face_detector = facer.face_detector("retinaface/mobilenet", device=device)
with torch.inference_mode():
    faces = face_detector(image)

face_attr = facer.face_attr("farl/celeba/224", device=device)
with torch.inference_mode():
    faces = face_attr(image, faces)

labels = face_attr.labels
face1_attrs = faces["attrs"][0] # get the first face's attributes

print(labels)

for prob, label in zip(face1_attrs, labels):
    if prob > 0.5:
        print(label, prob.item())

Check this notebook for full example.

Please consider citing

@inproceedings{zheng2022farl,
  title={General facial representation learning in a visual-linguistic manner},
  author={Zheng, Yinglin and Yang, Hao and Zhang, Ting and Bao, Jianmin and Chen, Dongdong and Huang, Yangyu and Yuan, Lu and Chen, Dong and Zeng, Ming and Wen, Fang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={18697--18709},
  year={2022}
}

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distribution

pyfacer-0.0.5-py3-none-any.whl (39.4 kB view details)

Uploaded Python 3

File details

Details for the file pyfacer-0.0.5-py3-none-any.whl.

File metadata

  • Download URL: pyfacer-0.0.5-py3-none-any.whl
  • Upload date:
  • Size: 39.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.0.1 CPython/3.10.14

File hashes

Hashes for pyfacer-0.0.5-py3-none-any.whl
Algorithm Hash digest
SHA256 8cc21b3f3c7be3438e81072f444fd947f1bea78fa8919d6c6f1b0527690ac42d
MD5 6e5143ac941799b2a9aab488b8d2b57b
BLAKE2b-256 5cc61afa21f64d137d01558a0492ab641d58412f6c87e84a79753b30c93d97f1

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page