Fingerprint dataset collection

Curated collection of human fingerprint datasets suitable for research and evaluation of fingerprint recognition algorithms.

• Dataset classification
• Public rectangular datasets
  • FVC2000 DB1 B
  • FVC2000 DB2 B
  • FVC2000 DB3 B
  • FVC2000 DB4 B
  • FVC2002 DB1 B
  • FVC2002 DB2 B
  • FVC2002 DB3 B
  • FVC2002 DB4 B
  • FVC2004 DB1 B
  • FVC2004 DB2 B
  • FVC2004 DB3 B
  • FVC2004 DB4 B
  • Neurotechnology CrossMatch
  • Neurotechnology UareU
  • FVS dataset
  • Tsinghua Palmprint Database
  • COEP Palm Print Database
  • Birjand University Mobile Palmprint Database
• Licensed rectangular datasets
  • CASIA-FingerprintV5
  • NIST Special Database 302
  • NIST Special Database 301 A
  • FVC2006 DB1
  • FVC2006 DB2
  • FVC2006 DB3
  • FVC2006 DB4
  • CASIA-PalmprintV1
  • CASIA-MS-PalmprintV1
  • FVC2000 DB1 A
  • FVC2000 DB2 A
  • FVC2000 DB3 A
  • FVC2000 DB4 A
  • FVC2002 DB1 A
  • FVC2002 DB2 A
  • FVC2002 DB3 A
  • FVC2002 DB4 A
  • FVC2004 DB1 A
  • FVC2004 DB2 A
  • FVC2004 DB3 A
  • FVC2004 DB4 A
• Public pairsets
  • MINEX validation dataset
• Licensed pairsets
  • NIST Special Database 300
• Licensed latent datasets
  • NIST Special Database 302 E
  • NIST Special Database 301 B
• Public unpaired datasets
  • SOCOFing
• Secret datasets
• Generators
• Other lists

Dataset classification

By access:

• public - Available for download by anyone without any explicit restrictions on use or redistribution. Where implicit copyright applies, publisher appears to be uninterested in enforcing it.
• licensed - Available only after accepting license that contains confidentiality terms. This includes all datasets that cost money. License may include other restrictions in addition to confidentiality.
• secret - Secret dataset itself cannot be obtained, but algorithms can be submitted for evaluation on the dataset. All competition datasets are secret to prevent cheating.

By impression count:

• rectangular dataset - There are more than two impressions per finger. Rectangular datasets have certain advantages in research due to the large number of matching pairs.
• pairset - There are two impressions per finger. All natural datasets and thus all large datasets are pairsets.
• latent dataset - Latent datasets contain latent fingerprints taken from objects. They typically do not identify finger, only subject. They are usually intended to be matched against plain/rolled fingerprints from another dataset.
• unpaired - Unpaired datasets contain only one impression of every finger. They have somewhat limited applications.

Public rectangular datasets

FVC2000 DB1 B

Small sample of FVC2000 DB1, one of the four datasets used in FVC2000 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 300x300px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2000 datasets
• Population: white students
• Enrollment: unmoderated
• Sensor: optical, 500dpi, Secure Desktop Scanner by KeyTronic

FVC2000 DB2 B

Small sample of FVC2000 DB2, one of the four datasets used in FVC2000 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 256x364px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2000 datasets
• Population: white students
• Enrollment: unmoderated
• Sensor: capacitive, 500dpi, TouchChip by ST Microelectronics

FVC2000 DB3 B

Small sample of FVC2000 DB3, one of the four datasets used in FVC2000 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 448x478px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2000 datasets
• Population: white, 5-73 years
• Enrollment: partial QA
• Sensor: optical, 500dpi, DF-90 by Identicator Technology

FVC2000 DB4 B

Small sample of FVC2000 DB4, one of the four datasets used in FVC2000 competition. This dataset is synthetic.

• Size: 10 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 240x320px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2000 datasets
• Generator: unknown

FVC2002 DB1 B

Small sample of FVC2002 DB1, one of the four datasets used in FVC2002 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 388x374px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 500dpi, TouchView II by Identix

FVC2002 DB2 B

Small sample of FVC2002 DB2, one of the four datasets used in FVC2002 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 569dpi, 296x560px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 569dpi, FX2000 by Biometrika

FVC2002 DB3 B

Small sample of FVC2002 DB3, one of the four datasets used in FVC2002 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 300x300px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: capacitive, 500dpi, 100 SC by Precise Biometrics

FVC2002 DB4 B

Small sample of FVC2002 DB4, one of the four datasets used in FVC2002 competition. This dataset is synthetic.

• Size: 10 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 288x384px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2002 datasets
• Generator: SFinGe v2.51

FVC2004 DB1 B

Small sample of FVC2004 DB1, one of the four datasets used in FVC2004 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 640x480px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 500dpi, V300 by CrossMatch

FVC2004 DB2 B

Small sample of FVC2004 DB2, one of the four datasets used in FVC2004 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 512dpi, 328x364px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 512dpi, U.are.U 4000 by Digital Persona

FVC2004 DB3 B

Small sample of FVC2004 DB3, one of the four datasets used in FVC2004 competition.

• Size: 10 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 512dpi, 300x480px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: thermal sweep, 512dpi, FingerChip FCD4B14CB by Atmel

FVC2004 DB4 B

Small sample of FVC2004 DB4, one of the four datasets used in FVC2004 competition. This dataset is synthetic.

• Size: 10 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 288x384px
• License: unspecified, free download
• Download: download page, direct download
• Documentation: FVC2004 datasets
• Generator: SFinGe v3.0

Neurotechnology CrossMatch

One of two sample datasets distributed by Neurotechnology.

• Size: 51 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 504x480px
• License: unspecified, free download
• Download: download page, direct download
• Sensor: optical, 500dpi, Verifier 300 by Cross Match Technologies

Neurotechnology UareU

One of two sample datasets distributed by Neurotechnology.

• Size: 65 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 512dpi, 326x357px
• License: unspecified, free download
• Download: download page, direct download
• Sensor: optical, 512dpi, U.are.U 4000 by DigitalPersona

FVS dataset

Small rectangular dataset shared by Shivang Patel as part of his Fingerprint Verification System project.

• Size: 21 fingers x 8 impressions
• Impression type: plain
• Format: BMP, 256x256px, unknown DPI
• License: unspecified, free download
• Download: download page, direct download

Tsinghua Palmprint Database

Tsinghua Palmprint Database (THUPALMLAB) contains palm prints without fingers.

• Size: 80 subjects x 2 palms x 8 impressions
• Impression type: palm
• Format: BMP, 500dpi, 2040x2040px
• License: for noncommercial research
• Download: download page, direct download
• Documentation: paper
• Sensor: Hisign

COEP Palm Print Database

Contactless right hand photos. Friction ridges are rarely visible.

• Size: 168 subjects x 1 hand x 8 impressions
• Impression type: contactless right hand, usually without friction ridges
• Format: color JPEG, 1600x1200px, narrow DPI range
• License: free download
• Download: download page, direct download
• Documentation: homepage
• Population: citizens of India
• Enrollment: pegs fix hand position
• Sensor: digital camera

Birjand University Mobile Palmprint Database

Contactless hand and palm photos from University of Birjand. Friction ridges are rarely visible.

• Size: 41 subjects x 2 hands x 20 impressions
• Impression type: contactless hand and palm, usually without friction ridges
• Format: color JPEG, up to 2448x3264px, narrow DPI range
• License: unknown
• Download: Kaggle
• Documentation: Kaggle page
• Population: Iranian women
• Enrollment: encoraged variation
• Sensor: digital camera

Licensed rectangular datasets

CASIA-FingerprintV5

Large dataset collected by Chinese Academy of Sciences.

• Size: 500 subjects x 8 fingers x 5 impressions
• Impression type: plain
• Format: BMP, 512dpi, 328x356px
• License: bundled in dataset description, prohibits publishing and redistribution
• Download: download page (requires signup)
• Documentation: attached on download page
• Population: Chinese graduate students, workers, waiters, ...
• Enrollment: artificial difficulty
• Sensor: "URU4000" (probably U.are.U 4000 by Digital Persona, optical, 512dpi)

NIST Special Database 302

Live fingerprint dataset collected by NIST using a wide variety of sensors. It also includes latent fingerprints that have their own entry in this list.

• Size: 200 subjects x 10 fingers x 12-18 impressions
• Impression type: plain, rolled, 4-finger, 2-thumb, palm
• Format: PNG, 500-1000dpi
• License: simple confidentiality rules
• Download: via request page
• Documentation: overview page, user guide
• Population: volunteers from US
• Enrollment: moderated
• Sensor: 15 sensor types (7 optical, 3 solid-state, 5 touch-free)

NIST Special Database 301 A

Live fingerprint dataset collected by NIST using a wide variety of sensors.

• Size: 51 subjects x 10 fingers x 14-15 impressions
• Impression type: plain, rolled, 4-finger, 2-thumb, palm
• Format: PNG, 500-1000dpi
• License: simple confidentiality rules
• Download: via request page
• Documentation: overview page, user guide
• Population: volunteers from US
• Enrollment: moderated
• Sensor: 10 sensor types (5 optical, 2 solid-state, 3 touch-free)

FVC2006 DB1

One of the four datasets used in FVC2006 competition. It is split into A (140 fingers) and B (10 fingers) subsets.

• Size: 150 fingers x 12 impressions
• Impression type: plain
• Format: BMP, 250dpi, 96x96px
• License: signed 2-year license agreement
• Download: see instructions
• Documentation: FVC2006 datasets
• Population: white volunteers
• Enrollment: unmoderated
• Sensor: electric field sensor, 250dpi

FVC2006 DB2

One of the four datasets used in FVC2006 competition. It is split into A (140 fingers) and B (10 fingers) subsets.

• Size: 150 fingers x 12 impressions
• Impression type: plain
• Format: BMP, 569dpi, 400x560px
• License: signed 2-year license agreement
• Download: see instructions
• Documentation: FVC2006 datasets
• Population: white volunteers
• Enrollment: unmoderated
• Sensor: electric field sensor, 250dpi

FVC2006 DB3

One of the four datasets used in FVC2006 competition. It is split into A (140 fingers) and B (10 fingers) subsets.

• Size: 150 fingers x 12 impressions
• Impression type: sweep
• Format: BMP, 500dpi, 400x500px
• License: signed 2-year license agreement
• Download: see instructions
• Documentation: FVC2006 datasets
• Population: white volunteers
• Enrollment: unmoderated
• Sensor: electric field sensor, 250dpi

FVC2006 DB4

One of the four datasets used in FVC2006 competition. It is split into A (140 fingers) and B (10 fingers) subsets.

• Size: 150 fingers x 12 impressions
• Impression type: synthetic plain
• Format: BMP, 500dpi, 288x384px
• License: signed 2-year license agreement
• Download: see instructions
• Documentation: FVC2006 datasets
• Generator: SFinGe v3.0

CASIA-PalmprintV1

Palmprint dataset collected by Chinese Academy of Sciences. No friction ridges.

• Size: 312 subjects x 2 hands x 8 impressions
• Impression type: contactless palm without friction ridges
• Format: grayscale JPEG, 640x480px, variable DPI
• License: bundled in dataset description, prohibits publishing and redistribution
• Download: download page (requires signup)
• Documentation: attached on download page
• Population: Chinese citizens
• Sensor: custom

CASIA-MS-PalmprintV1

Multispectral hand dataset collected by Chinese Academy of Sciences. No friction ridges.

• Size: 100 subjects x 2 hands x 6 impressions x 6 wavelengths
• Impression type: contactless hand without friction ridges
• Format: grayscale JPEG, 768x576px, variable DPI
• License: bundled in dataset description, prohibits publishing and redistribution
• Download: download page (requires signup)
• Documentation: attached on download page
• Population: Chinese citizens
• Enrollment: pegs control hand position
• Sensor: custom

FVC2000 DB1 A

The larger, hidden part of FVC2000 DB1, one of the four datasets used in FVC2000 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 300x300px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2000 datasets
• Population: white students
• Enrollment: unmoderated
• Sensor: optical, 500dpi, Secure Desktop Scanner by KeyTronic

FVC2000 DB2 A

The larger, hidden part of FVC2000 DB2, one of the four datasets used in FVC2000 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 256x364px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2000 datasets
• Population: white students
• Enrollment: unmoderated
• Sensor: capacitive, 500dpi, TouchChip by ST Microelectronics

FVC2000 DB3 A

The larger, hidden part of FVC2000 DB3, one of the four datasets used in FVC2000 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 448x478px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2000 datasets
• Population: white, 5-73 years
• Enrollment: partial QA
• Sensor: optical, 500dpi, DF-90 by Identicator Technology

FVC2000 DB4 A

The larger, hidden part of FVC2000 DB4, one of the four datasets used in FVC2000 competition. This dataset is synthetic.

• Size: 100 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 240x320px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2000 datasets
• Generator: unknown

FVC2002 DB1 A

The larger, hidden part of FVC2002 DB1, one of the four datasets used in FVC2002 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 388x374px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 500dpi, TouchView II by Identix

FVC2002 DB2 A

The larger, hidden part of FVC2002 DB2, one of the four datasets used in FVC2002 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 569dpi, 296x560px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 569dpi, FX2000 by Biometrika

FVC2002 DB3 A

The larger, hidden part of FVC2002 DB3, one of the four datasets used in FVC2002 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 300x300px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2002 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: capacitive, 500dpi, 100 SC by Precise Biometrics

FVC2002 DB4 A

The larger, hidden part of FVC2002 DB4, one of the four datasets used in FVC2002 competition. This dataset is synthetic.

• Size: 100 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 288x384px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2002 datasets
• Generator: SFinGe v2.51

FVC2004 DB1 A

The larger, hidden part of FVC2004 DB1, one of the four datasets used in FVC2004 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 500dpi, 640x480px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 500dpi, V300 by CrossMatch

FVC2004 DB2 A

The larger, hidden part of FVC2004 DB2, one of the four datasets used in FVC2004 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 512dpi, 328x364px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: optical, 512dpi, U.are.U 4000 by Digital Persona

FVC2004 DB3 A

The larger, hidden part of FVC2004 DB3, one of the four datasets used in FVC2004 competition.

• Size: 100 fingers x 8 impressions
• Impression type: plain
• Format: TIFF, 512dpi, 300x480px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2004 datasets
• Population: white students
• Enrollment: artificial difficulty
• Sensor: thermal sweep, 512dpi, FingerChip FCD4B14CB by Atmel

FVC2004 DB4 A

The larger, hidden part of FVC2004 DB4, one of the four datasets used in FVC2004 competition. This dataset is synthetic.

• Size: 100 fingers x 8 impressions
• Impression type: synthetic plain
• Format: TIFF, 500dpi, 288x384px
• Cost: ~140€ (includes all 12 FVC datasets and a book)
• Documentation: FVC2004 datasets
• Generator: SFinGe v3.0

Public pairsets

MINEX validation dataset

Small public dataset that can be used to prepare submissions to MINEX competition. Part of MINEX validation tool.

• Size: 50 subjects x 10 fingers x 2 impressions
• Impression type: plain
• Format: raw grayscale, separate metadata, 500dpi
• License: permissive
• Download: GitHub

Licensed pairsets

NIST Special Database 300

Segmented scanned fingerprint cards from US law enforcement. This is technically a single impression dataset, but since every finger is present twice as rolled and flat impression, the dataset can be considered a pairset.

• Size: 888 subjects x 10 fingers x 2 impressions
• Impression type: plain, rolled, 4-finger
• Format: PNG, 500/1000/2000dpi
• License: simple confidentiality rules
• Download: via request page
• Documentation: overview page, user guide
• Population: arrested US citizens
• Enrollment: controlled but with some uncooperative subjects
• Technology: scanned ink

Licensed latent datasets

NIST Special Database 302 E

Dataset of latent fingerprints of subjects from NIST Special Database 302. Latent fingerprints from this dataset are supposed to be matched with plain/rolled fingerprints from NIST Special Database 302.

• Size: 200 subjects x 50 impressions
• Impression type: latent, finger and palm
• Format: PNG, 1000-1500dpi
• License: simple confidentiality rules
• Download: via request page
• Documentation: overview page, user guide
• Population: volunteers from US, subjects identical to NIST Special Database 302
• Activity: variety of activities that result in latent fingerprints
• Technology: variety of latent fingerprint collection methods

NIST Special Database 301 B

Dataset of latent fingerprints of subjects from NIST Special Database 301 A. Latent fingerprints from this dataset are supposed to be matched with plain/rolled fingerprints from NIST Special Database 301 A.

• Size: 24 subjects x 50 impressions
• Impression type: latent, finger and palm
• Format: PNG, 1000-1500dpi
• License: simple confidentiality rules
• Download: via request page
• Documentation: overview page, user guide
• Population: volunteers from US, subset of subjects from NIST Special Database 301 A
• Activity: variety of activities that result in latent fingerprints
• Technology: variety of latent fingerprint collection methods

Public unpaired datasets

SOCOFing

Sokoto Coventry Fingerprint Dataset (SOCOFing) is a single impression dataset. Images are tagged with gender and finger position. Real images are complemented with synthetically damaged versions.

• Size: 600 subjects x 10 fingers x 1 impression
• Impression type: plain
• Format: BMP, 500dpi, 96x103px
• License: for noncommercial research, see paper
• Download: Kaggle
• Documentation: Arxiv paper
• Population: African
• Sensor: Hamster Plus (HSDU03P), SecuGen SDU03P

Secret datasets

Datasets of some of these competitions might overlap. There's no way to know for sure, so all competitions are listed.

• FVC-onGoing Fingerprint Verification (FV)
• FVC-onGoing Palmprint Verification (PV)
• FVC-onGoing Fingerprint Matching (FMISO)
• FVC-onGoing Fingerprint Indexing (FIDX)
• FVC-onGoing Fingerprint Orientation Extraction (FOE)
• FVC-onGoing Secure Template Fingerprint Verification (STFV)
• NIST Minutiae Interoperability Exchange (MINEX)
• NIST Slap Fingerprint Segmentation Evaluation (SlapSeg)
• NIST Proprietary Fingerprint Template (PFT)
• NIST Evaluation of Latent Friction Ridge Technology (ELFT)

Generators

• SFinGe

Other lists

• ⭐ NIST Biometric and Forensic Research Database Catalog
• Luigi Rosa's list
• CVonline
• NIST Special Databases
• FVC2000, FVC2002, FVC2004

Contribute

If you have something to add or correct here, please submit a PR or open an issue.

Legal

Every dataset has its own license. This CC0 list was created by Robert Važan while working on SourceAFIS.