Dataset Information

Pgu-Face: A dataset of partially covered facial images.

ABSTRACT: In this article we introduce a human face image dataset. Images were taken in close to real-world conditions using several cameras, often mobile phone?s cameras. The dataset contains 224 subjects imaged under four different figures (a nearly clean-shaven countenance, a nearly clean-shaven countenance with sunglasses, an unshaven or stubble face countenance, an unshaven or stubble face countenance with sunglasses) in up to two recording sessions. Existence of partially covered face images in this dataset could reveal the robustness and efficiency of several facial image processing algorithms. In this work we present the dataset and explain the recording method.

SUBMITTER: Salari SR

PROVIDER: S-EPMC5024140 | biostudies-literature | 2016 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Pgu-Face: A dataset of partially covered facial images.

Salari Seyed Reza SR Rostami Habib H

Data in brief 20160910

In this article we introduce a human face image dataset. Images were taken in close to real-world conditions using several cameras, often mobile phone׳s cameras. The dataset contains 224 subjects imaged under four different figures (a nearly clean-shaven countenance, a nearly clean-shaven countenance with sunglasses, an unshaven or stubble face countenance, an unshaven or stubble face countenance with sunglasses) in up to two recording sessions. Existence of partially covered face images in this ...[more]

PMID: 27668275

Similar Datasets

Project description:BACKGROUND:Patients with facial nerve paralysis (FNP) experience challenges in accessing health care that could potentially be overcome by telemedicine. However, the reliability of telemedicine has yet to be established in this field. OBJECTIVE:This study aimed to investigate the consistency between face-to-face and video assessments of patients with FNP by experienced clinicians. METHODS:A repeated-measures design was used. A total of 7 clinicians assessed the FNP of 28 patients in a face-to-face clinic using standardized grading systems (the House-Brackmann, Sydney, and Sunnybrook facial grading systems). After 3 months, the same grading systems were used to assess facial palsy in video recordings of the same patients. RESULTS:The House-Brackmann system in video assessment had excellent reliability and agreement (intraclass correlation coefficient [ICC]=0.780; principal component analysis [PCA]=87.5%), similar to face-to-face assessment (ICC=0.686; PCA=79.2%). Reliability of the Sydney system was good to excellent, with excellent agreement face-to-face (ICC=0.633 to 0.834; PCA=81.0%-95.2%). However, video assessment of the cervical branch and synkinesis had fair reliability and good agreement (ICC=0.437 to 0.597; PCA=71.4%), whereas that of other branches had good to excellent reliability and excellent agreement (ICC=0.625 to 0.862; PCA=85.7%-100.0%). Reliability of the Sunnybrook system was poor to fair for resting symmetry (ICC=0.195 to 0.498; PCA=91.3%-100.0%) and synkinesis (ICC=-0.037 to 0.637; PCA=69.6%-87.0%) but was good to excellent for voluntary movement (ICC=0.601 to 0.906; PCA=56.5%-91.3%) in face-to-face and video assessments. Bland-Altman plots indicated normal limits of agreement within ±1 between face-to-face and video-assessed scores only for the temporal and buccal branches of the Sydney system and for resting symmetry in the Sunnybrook system. CONCLUSIONS:Video assessment of FNP with the House-Brackmann and Sunnybrook systems was as reliable as face-to-face but with insufficient agreement, especially in the assessment of synkinesis. However, video assessment does not account for the impact of real-time interactions that occur during tele-assessment sessions.

Dataset Information

Pgu-Face: A dataset of partially covered facial images.

Publications

Pgu-Face: A dataset of partially covered facial images.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure