Palmprint Database (NUIG_Palm1)

Palmprint Database for palmprint recognition using smartphones in unconstrained conditions

This database is used in the paper “Unconstrained palmprint as a smartphone biometric” published in IEEE Transactions on Consumer Electronics, vol. 63, no. 3, pp. 334–342, Aug. 2017.

Main Database Characteristics

Upper row: Sample images from the database; Lower row: Sample ROIs extracted and normalized

The proposed database provides researchers with a rich set of images taken from 81 individual subjects of mixed gender, and ages ranging from 19 to 55 years old. A total of 5 smartphone devices were used to acquire images for each subject. Images of each user’s palmprint were acquired under two distinct lighting levels and in two distinct background conditions. There are 20 images of palmprints per subject or 1,616 images available in the main database.

Note that only one hand was used per subject as the workflow reflected a user’s natural predilection to use their lead-hand to hold the device, thus capturing an image of their secondary hand. In order to increase the relevance of the database, all right hand images were flipped vertically so that all samples can be considered as coming from a left hand.
When downloading the dataset, all images (originals and vertically flipped) are included.

Device model Device# Sensor size Stabilization Sensor resolution Aperture Month of launch
LG G4 Device1 1/2.6″ Yes, 3 axes 16 MP f/1.8 April 2015
Samsung Galaxy S6 Device2 1/2.6″ Yes, 3 axes 16 MP f/1.9 April 2015
Apple iPhone 5 Device3 1/3.2″ No 8 MP f/2.4 Sept. 2012
Apple iPhone 6S Device4 1/3″ Yes, 1 axis 12 MP f/2.2 Sept. 2015
Huawei P8 Device5 1/3.06″ Yes, 1 axis 13 MP f/2.0 April 2015


Extracted Regions of Interest (ROIs)

Region of Interest (ROI) extraction visualized

A generic processing pipeline requires the palmprint to be extracted from the hand in a consistent manner.  The finger bases were marked manually with 5 points, where the 3rd one marks the central finger valley. If we denote the first two points X1, X2 and the last two points as X4 and X5, then the middle point of these segments are represented by X12 and X45. They are then used to create a new coordinate system to rotate and align the palmprints, as demonstrated in the figure above, where the extracted ROI is contained within the black square. These landmarks and the extracted ROIs are provided as benchmarks in future tests related to hand detection.

Access to the database

If you are interested in downloading the database, please send an e-mail to or and we will arrange for the signing of a License Agreement. Temporary access of up to 1 week will be then offered to download the images and the associated XML files from our server.


If you use the database in your work and it gets published, please cite our database using the article it was first used in:

@article{author = {Ungureanu, Adrian-Stefan and Thavalengal, Shejin and Cognard, Timoth{\'{e}}e E. and Costache, Claudia and Corcoran, Peter}, doi = {10.1109/TCE.2017.014994}, isbn = {9781467383646}, issn = {0098-3063}, journal = {IEEE Transactions on Consumer Electronics}, keywords = {Biometrics,consumer electronics,security}, month = {aug}, number = {3}, pages = {334–342}, title = {{Unconstrained palmprint as a smartphone biometric}}, url = {}, volume = {63}, year = {2017}