| Deepfake-Eval-2024 |
Deepfake-Eval-2024 consists of 45 hours of videos, 56.5 hours of audio, and 1,975 images. |
2025 |
Access Link |
Cite |
TrueMedia.org |
| AI-Face |
It’s the first million-scale demographically annotated AI-generated face image dataset. |
2025 |
Access Link |
Cite |
Purdue University |
| UniAttackData |
This dataset contains 29,706 videos from 1,800 participants, covering 12 types of digital attacks and 2 types of physical attacks. |
2024 |
Access Link |
Cite |
CASIA |
| SuHiFiMask |
It includes 10,195 video clips, covering 232 types of 3D silicone masks, 200 variants of 2D printed/screen attacks, and adversarial samples. |
2023 |
Access Link |
Cite |
CASIA |
| SiW-Mv2 |
SiW-Mv2 collects 785 videos from 493 subjects, and 915 spoof videos from 600 subjects. |
2022 |
Access Link |
Cite |
Michigan State University |
| PADISI-Face |
It encompasses physical, digital, and combined attack types, and provides interpretability annotations such as natural language descriptions. |
2021 |
Access Link |
Cite |
USC Information Sciences Institute |
| Glint360K |
Glint360K contains 17091657 images of 360232 individuals. |
2020 |
Access Link |
Cite |
InsightFace / DeepGlint |
| Tufts Face Database |
The Tufts-Face-Database contains over 100,000 multi-modal face images, encompassing seven image modalities. |
2020 |
Access Link |
Cite |
Tufts University |
| HQ-WMCA |
It consists of 2904 short multi-modal video recordings of both bona-fide and presentation attacks. |
2020 |
Access Link |
Cite |
Idiap Research Institute |
| CelebA-Spoof |
It has 625,537 images from 10,177 subjects, which includes 43 rich attributes on face, illumination, environment and spoof types. |
2020 |
Access Link |
Cite |
Beijing Jiaotong University |
| Flickr-Faces-HQ Dataset (FFHQ) |
The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution. |
2019 |
Access Link |
Cite |
NVIDIA |
| WMCA |
1941 short video recordings of both bonafide and presentation attacks from 72 different identities. |
2019 |
Access Link |
Cite |
Idiap Research Institute |
| VGGFace2 |
VGGFace2 is a large-scale face recognition dataset containing over 3.3 million images of more than 9,000 identities. |
2018 |
Access Link |
Cite |
University of Oxford |
| FEC Google |
This dataset is composed of face image triplets and annotations. The annotations are integers from the set {1, 2, 3}. |
2018 |
Access Link |
Cite |
Google |
| Face Images with Marked Landmark Points |
The dataset contains 7049 facial images, with a maximum of 15 keypoints marked on these images. |
2018 |
Access Link |
Cite |
Ball State University |
| CSMAD |
The Custom Silicone Mask Attack Dataset (CSMAD) contains presentation attacks made of six custom-made silicone masks. Each mask cost about USD 4000. |
2018 |
Access Link |
Cite |
Idiap Research Institute |
| SiW |
It has total 4,478 videos. All videos are in 30 fps, about 15 second length, and 1080P HD resolution. |
2018 |
Access Link |
Cite |
Michigan State University |
| CPLFW |
CPLFW deliberately searches and selects 3,000 positive face pairs with pose difference to add pose variation to intra-class variance. |
2018 |
Access Link |
Cite |
BUPT |
| TinyFace |
The TinyFace dataset consists of 5,139 labelled facial identities given by 169,403 native LR face images. |
2018 |
Access Link |
Cite |
Queen Mary University of London |
| IJB-C |
IJB-C includes a total of 31, 334 (21, 294 face and 10, 040 non-face) still images, averaging to ~6 images per subject, and 117, 542 frames from 11, 779 full-motion videos, averaging to ~33 frames per subject and ~3 videos per subject. |
2018 |
Access Link |
Cite |
IARPA / NIST |
| CALFW |
CALFW deliberately searches and selects 3,000 positive face pairs with age gaps to add aging process intra-class variance. |
2017 |
Access Link |
Cite |
BUPT |
| UTKFace |
The UTKFace dataset consists of over 20,000 facial images. Each image is annotated with age, gender, and ethnicity information. |
2017 |
Access Link |
Cite |
The University of Tennessee |
| AgeDB |
AgeDB contains 16, 488 images of various famous people, such as actors/actresses, writers, scientists, politicians, etc. |
2017 |
Access Link |
Cite |
Imperial College London |
| OULU-NPU |
It consists of 4950 real access and attack videos. |
2017 |
Access Link |
Cite |
University of Oulu |
| CFP |
The data set contains 10 frontal and 4 profile images of 500 individuals. |
2016 |
Access Link |
Cite |
University of Maryland |
| MS-Celeb-1M |
MS-Celeb-1M is a dataset of 10 million face images harvested from the Internet. |
2016 |
Access Link |
Cite |
Microsoft Research |
| Wider Face |
Wider Face has more than 10,000 images of singles and groups of people. |
2016 |
Access Link |
Cite |
The Chinese University of Hong Kong |
| The Hong Kong Polytechnic University Disguise and Makeup Faces Database |
This database consists of 2460 different face images from 410 different subjects. |
2016 |
Access Link |
Cite |
Department of Computing, The Hong Kong Polytechnic University |
| CelebA |
The CelebA dataset contains over 200,000 celebrity images, covering significant pose variations and cluttered backgrounds. |
2015 |
Access Link |
Cite |
The Chinese University of Hong Kong |
| MALF |
It contains 5,250 images and 11,931 annotated faces collected from the Internet. |
2015 |
Access Link |
Cite |
CASIA |
| CASIA-WebFace |
The dataset contains 494,414 face images of 10,575 real identities collected from the web. |
2014 |
Access Link |
Cite |
CASIA |
| MSU MFSD |
The MSU MFSD database consists of 440 video clips of photo and video attack attempts of 55 subjects. |
2014 |
Access Link |
Cite |
Michigan State University |
| 3DMAD |
It contains 76500 frames of 17 persons. Each frame consists of a depth image (640×480 pixels – 1×11 bits) |
2013 |
Access Link |
Cite |
Izmir Institute of Technology |
| ReplayAttack |
It consists of 1300 video clips of photo and video attack attempts to 50 clients. |
2012 |
Access Link |
Cite |
Idiap Research Institute |
| YouTube Faces With Facial Keypoints |
It contains short videos of celebrities downloaded from YouTube, with a maximum of 6 videos per celebrity. |
2011 |
Access Link |
Cite |
Tel-Aviv University |
| CASIA-FaceV5 |
This dataset comprises 2,500 high-resolution color images (640×480 pixels in BMP format). |
2010 |
Access Link |
Cite |
CASIA |
| CASIA-3D-FaceV1 |
It comprises 123 scans, with each scan simultaneously generating a paired 2D color image and a 3D facial triangular mesh surface. |
2008 |
Access Link |
Cite |
CASIA |
| IIT Delhi Near IR Face Database (Version 2.0) |
The database of 612 images is organized into two folders (each with 306 images from 102 users). |
2008 |
Access Link |
Cite |
Indian Institute of Technology Delhi |
| Labeled Faces in the Wild |
It contains 13,233 images of 5,749 individuals, among whom 1,680 people have two or more distinct photos. |
2007 |
Access Link |
Cite |
University of Massachusetts Amherst |
| MORPH |
MORPH is a facial age estimation dataset, which contains 55,134 facial images of 13,617 subjects ranging from 16 to 77 years old. |
2006 |
Access Link |
Cite |
University of North Carolina Wilmington |