Identity Similarity

This repository can help researchers who want to use face recognition in their research. You can easily implement powerful face recognition models in your project. I motivated by LPIPS for this repository. The models are borrowed from Insigtface.

Warning : Please, be careful when chosing your criterion. Lower is more similar in MSE while higher is more similar in CosineSimilarity.

Supported Metrics

• MSE
• L1
• Cosine

Supported Models

• ms1mv3_arcface_r50
• glint360k_cosface_r100

Usage

1. Training with preprocessed dataset.

In this case, we assume that you have aligned images using a keypoint template and you want to calculate identity similarity between two aligned images or a image and a saved identity vector.

import torch
import numpy as np
from idsim import IdentitySimilarity

idsim = IdentitySimilarity()
template = np.array([[35.066223, 34.23266],
                  [84.1586, 33.96113],
                  [59.768444, 62.152763],
                  [39.60066, 90.89288],
                  [80.255, 90.66802]], dtype=np.float32)
idsim.set_ref_point(template)

# dummy variables
v1 = torch.rand(1, 512)
im1 = torch.rand(5, 3, 128, 128)

# useful functions
sim_v2v = idsim.forward_v2v(v1, v1)
sim_im2im = idsim.forward_img2img(im1, im1)
sim_v2im = idsim.forward_v2img(v1, im1)
print("\nsim_v2v :", sim_v2v, "\nsim_im2im :", sim_im2im, "\nsim_v2im :", sim_v2im)

2. Face recognition

In this case, Idsim can caculate identity similarity of your images.

import cv2
from idsim import IdentitySimilarity

idsim = IdentitySimilarity(criterion="Cosine")
img1 = cv2.imread("a.jpg")
img2 = cv2.imread("b.jpg")
v1 = idsim.extract_identity(img1) 
v2 = idsim.extract_identity(img2)
sim = idsim.forward_v2v(v1,v2)
print("Similarity :", sim)

Note: You can check the proving_differentiability.ipynb for an example training.

Contributing

Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Run make style && make quality in the root repo directory, to ensure code quality.
4. Commit your Changes (git commit -m 'Add some AmazingFeature')
5. Push to the Branch (git push origin feature/AmazingFeature)
6. Open a Pull Request