All Projects → seriousran → face-recognition

seriousran / face-recognition

Licence: other
얼굴 인식에 대한 기술 동향 및 관련 모델 자료

Projects that are alternatives of or similar to face-recognition

FaceIDLight
A lightweight face-recognition toolbox and pipeline based on tensorflow-lite
Stars: ✭ 17 (-55.26%)
Mutual labels:  face, face-recognition, face-identification
Pychubby
Automated face warping tool.
Stars: ✭ 240 (+531.58%)
Mutual labels:  face, face-recognition
Faceimagequality
Code and information for face image quality assessment with SER-FIQ
Stars: ✭ 223 (+486.84%)
Mutual labels:  face, face-recognition
Bird Recognition Review
A list of useful resources in the bird sound (song and calls) recognition, such as datasets, papers, links to open source projects and competitions
Stars: ✭ 116 (+205.26%)
Mutual labels:  review, survey
Guided-I2I-Translation-Papers
Guided Image-to-Image Translation Papers
Stars: ✭ 117 (+207.89%)
Mutual labels:  cvpr, iccv
Lookatme
VideoView that plays video only when 👀 are open and 👦 is detected with various other features
Stars: ✭ 161 (+323.68%)
Mutual labels:  face, face-recognition
Data Augmentation Review
List of useful data augmentation resources. You will find here some not common techniques, libraries, links to github repos, papers and others.
Stars: ✭ 785 (+1965.79%)
Mutual labels:  review, survey
Awesome Face
😎 face releated algorithm, dataset and paper
Stars: ✭ 739 (+1844.74%)
Mutual labels:  face, face-recognition
AIPaperCompleteDownload
Complete download for papers in various top conferences
Stars: ✭ 64 (+68.42%)
Mutual labels:  cvpr, iccv
Nearest-Celebrity-Face
Tensorflow Implementation of FaceNet: A Unified Embedding for Face Recognition and Clustering to find the celebrity whose face matches the closest to yours.
Stars: ✭ 30 (-21.05%)
Mutual labels:  face, face-recognition
Face Recognition.js
Simple Node.js package for robust face detection and face recognition. JavaScript and TypeScript API.
Stars: ✭ 1,768 (+4552.63%)
Mutual labels:  face, face-recognition
facematch
Facematch is a tool to verifies if two photos contain the same person.
Stars: ✭ 62 (+63.16%)
Mutual labels:  face, face-recognition
Maad Face
A face database with a large number of high-quality attribute annotations
Stars: ✭ 112 (+194.74%)
Mutual labels:  face, face-recognition
Dockerface
Face detection using deep learning.
Stars: ✭ 173 (+355.26%)
Mutual labels:  face, face-recognition
Human Detection And Tracking
Human-detection-and-Tracking
Stars: ✭ 753 (+1881.58%)
Mutual labels:  face, face-recognition
survey-computer-vision-2021
2021年计算机视觉技术综述分类汇总
Stars: ✭ 54 (+42.11%)
Mutual labels:  survey, face-recognition
Facecropper
✂️ Crop faces, inside of your image, with iOS 11 Vision api.
Stars: ✭ 479 (+1160.53%)
Mutual labels:  face, face-recognition
Mobileface
A face recognition solution on mobile device.
Stars: ✭ 669 (+1660.53%)
Mutual labels:  face, face-recognition
Survey Computer Vision
2020-2021年计算机视觉综述论文分方向整理
Stars: ✭ 207 (+444.74%)
Mutual labels:  review, survey
FaceNet-IOT
IOT implementation for FaceNet project by David Sandberg https://github.com/davidsandberg/facenet
Stars: ✭ 18 (-52.63%)
Mutual labels:  face, face-recognition

face-recognition

_ Face Verification과 Identification을 모두 아우르는 얼굴 인식 기술에 대한 기술 동향_

ref: Mei Wang and Weihong Deng, "Deep Face Recognition: A Survey," 2018

Contents

논문 리뷰

Pre-Trained Model

Code

Dataset

Links

Face Recognition 서베이

ref: https://paperswithcode.com/

ref: https://paperswithcode.com/

2014 CVPR에서 대표적인 두 모델(DeepFace, DeepID)이 소개되면서 딥러닝 기법으로 활발히 연구가 시작된, 얼굴 인식 분야에 대해 최근 5년간의 동향을 다루고자 합니다.

  1. DeepFace (CVPR/2014)

    • Training Set: 4.4M
    • CNN Layers
    • LFW 97.35%
    • 3D Alignment
    • Ensembel Model
    • Softmax
    • ref:
  2. DeepID (CVPR/2014)

    • Training Set: 20W
    • CNN Layers
    • Softmax, Bayseian
    • LWF 97.45%
    • Softmax의 장점과 단점
      • Softmax는, Soft를 최대값으로 하는 것이 목표인 분류기입니다. CNN 분류 문제에서 매우 핫한 존재죠. 4가지를 분류한다고 가정하면, (x1, x2, x3, x4)의 값이 softmax 분류기에 의해 출력됩니다. 이것은 일반적으로 백분율 형태로 계산되고, 가장 큰 출력 값이 예상하는 값/결과가 됩니다.
      • Hard max vs Softmax
      • fig1~3
    • ref: Deep Learning Face Representation by Joint Identification-Verification
    • fig4
  3. FaceNet (CVPR/2015)

    • Softmax 대신 Triple Loss 사용
      • Triplet loss는 (a, p, n)의 triple 형태로 최적화합니다. 서로 다른(negative)의 특징의 L2거리가 유사한(positive) 특징의 L2 거리보다 크고, 클래스 간 간소화 및 클래스간 분리가 수행됨.
        • a: anchor
        • p: positive
        • n: negative
    • LFW: 99.64%
    • Training Set: 200M
    • use only 128 dim feature mapping
    • traiplet을 구현하는 코드는 매우 까다롭고 어려움
    • paper
    • 개념
    • triplet-loss
  4. Large Margin Softmax Loss (ICML/2016)

    • L-softmax는 Large Margin을 가진 softmax이다.
    • 큰 마진을 만드는 방법 = United Fully Conntected Layer + Softmax + Cross Entropy
    • Training Set: 0.49M
    • 17 CNN Layers
    • LFW: 98.72%
  5. sphereFace (CVPR/2017)

    • L-softmax를 개선하여 A-Softmax 제안
    • training sample 불균형의 수가 감소
    • mapping된 feature vector 각도의 최적화에 집중
    • Training Set: 0.49M
    • 64 CNN Layers
    • LFW: 99.42%
    • (현재 additive margin 시리즈를 훈련하는 것이 간단하고 성능도 더 좋음)
  6. Center Loss (ECCV/2016)

    • 카테고리에 대응하는 모든 feature vector에 대해 각 카테고리의 중심을 중앙으로 당김.
    • LFW(w/7 CNN Layers): 99.05
    • LFW(w/64 CNN Layers): 99.28
    • 다수의 클래스 센터를 유지하려면 메모리 소비가 비교적 큼.
  7. Center Invariant Loss (ACM MM/2017)

  8. Range Loss (ICCV/2017)

  9. Ring Loss (CVPR/2018)

  10. COCO | Congenerous Cosin (CVPR/2017)

  11. L2-constrained Softmax Loss (Arxiv/2017)

  12. NormFace (ACM MM/2017)

    • L2 HyperSphere Embedding
  13. AM-softmax (ICLR/2018)

    • Additive Margin Softmax
  14. CosFace (CVPR/2018)

  15. ArcFace (arXiv)

  16. InsightFace ()

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].