Sign in with the applicant's personal Hugging Face account, complete every required field below, and submit the request for manual review. / 请使用申请人本人的 Hugging Face 账号登录，完整填写下方必填字段并提交人工审核。

Access is available only to approved non-commercial academic projects. Applicants must have appropriate institutional ethics authorization and accept the PersoMoni Data Use Agreement. / 数据仅向审核通过的非商业学术研究项目开放。申请人须具备相应的机构伦理授权并接受 PersoMoni 数据使用协议。

PersoMoni Dataset

A Comprehensive Video-Based Benchmark Dataset for Fine-grained Personality Assessment with 15 Trait Dimensions

Language: English | 中文

How to apply

The application form is displayed in the Hugging Face access panel at the top of this dataset page. Sign in with the applicant's personal Hugging Face account, complete all identity, affiliation, research, ethics, and data-security fields, accept each required condition, and select Send request for manual review. Access decisions are tied to the individual account that submits the application.

Overview

PersoMoni is a video-based benchmark dataset for fine-grained personality assessment. Its BFI-2 annotations cover five broad personality domains and 15 facets, yielding 20 continuous personality targets. The dataset supports non-commercial academic research in video-based personality computing, multimodal learning, behavior analysis, and privacy-preserving representation learning.

Current release

The current version is an initial research release containing 5,000 processed facial clips selected with reference to the label distributions:

3,997 training clips and 1,003 test clips;
199,897 JPEG frames;
20 BFI-2 personality targets;
11 TAR data shards and five metadata/label files; and
pseudonymous sample identifiers only, with no original video identifiers, local paths, or private identity crosswalk.

This version does not contain the complete original interview videos and does not represent the full release of PersoMoni.

Dataset preview

Interactive Dataset Viewer previews are intentionally disabled because this release contains identifiable human-subject imagery. Approved applicants can download the original data shards and accompanying labels after manual access approval.

Future release plan

The project will continue feature-map extraction and quality control for the complete dataset. The full release is planned to focus on privacy-preserving feature representations. Its scope, licensing terms, and schedule will be announced after full-scale processing and quality assurance have been completed.

Access requirements

Applicants must provide:

a verifiable legal name, current affiliation, department, position, institutional email, country or region, and public academic profile;
the responsible principal investigator or supervisor and their institutional email;
a specific project title, research question, intended use, methods, evaluation plan, and expected outputs;
institutional ethics approval or documented exemption;
the proposed storage location, security controls, named team members, and requested access period; and
acceptance of every condition in the PersoMoni Data Use Agreement.

Every request is reviewed manually. Incomplete, commercial, identity-related, high-risk decision-making, or redistribution-oriented requests will not be approved. Granted access may be revoked when necessary.

Prohibited uses

Identifying, re-identifying, contacting, or linking participants to external data.
Face reconstruction, face recognition, biometric authentication, identity profiling, or surveillance.
High-stakes decisions involving employment, education, insurance, credit, healthcare, policing, or legal status.
Commercial use, sublicensing, publication of identifiable frames, or any redistribution of data or download credentials.
Inferring sensitive or protected attributes beyond the approved protocol.

PersoMoni Data Use Agreement — Draft v1.0

This version applies to the current research release. The data controller may require additional institutional documentation or a separately signed agreement before approval. By requesting or retaining access, the approved researcher ("Recipient") accepts the following terms.

Permitted purpose. Data may be used only by the named Recipient for the approved non-commercial academic project and during the approved period.
Named-user access. Access is personal. Unlisted students, collaborators, supervisors, services, and agents require separate approval.
No identification. The Recipient must not identify, re-identify, contact, track, or link participants to external information.
No biometric misuse. Face reconstruction, face recognition, biometric authentication, surveillance, and identity profiling are prohibited.
No consequential decisions. Data and derived models may not be used for high-stakes decisions affecting individuals.
No redistribution. Original files, excerpts, labels, pseudonymous IDs, credentials, and tokens may not be shared, published, sublicensed, sold, or uploaded to another service.
Secure storage. Data must remain on institutionally managed, access-controlled systems with encryption in transit and at rest, least-privilege access, and reasonable audit logging.
Public outputs. Publications may report aggregate statistics and non-identifying model results, but must not contain recognizable frames or outputs enabling reconstruction or participant linkage.
Incident reporting. Suspected loss, unauthorized access, disclosure, or misuse must be reported within 72 hours, followed by reasonable containment and investigation cooperation.
Deletion. All local, cloud, backup, and derived row-level copies must be deleted within 30 days after project completion, revocation, or termination unless retention is separately approved in writing.
Compliance. The Recipient is responsible for institutional ethics authorization and compliance with applicable privacy, research, and export laws.
Citation. Approved publications must cite the PersoMoni paper and acknowledge dataset access.
No warranty and revocation. Data are provided for research without warranty; access may be suspended or revoked to protect participants or ensure compliance.
Breach. A material violation requires immediate cessation of use and may lead to permanent rejection, institutional notification, and other remedies available to the data controller.

Submitting the application and selecting the agreement checkbox constitutes an electronic acknowledgment of these terms. The data controller may still require a separately signed document before approval.

Repository structure

data/
  train-00000-of-00008.tar
  ...
  test-00000-of-00003.tar
metadata/
  clips.csv
  frames.csv
  clip_labels.csv
  participant_labels.csv
  shard_checksums.csv
DATA_USE_AGREEMENT.md
ACCESS_REVIEW_CHECKLIST.md

TAR members use pseudonymous identifiers only, for example:

train/PMTR000001/frame_000.jpg

Citation

@article{cui2026persomoni,
  author  = {Cui, Feng-Qi and Huang, Jinyang and Zhao, Sirui and Li, Kun and Liu, Zhi and Li, Meng and Jia, Ziyu and Guo, Dan and Wang, Meng},
  title   = {PersoMoni: A Comprehensive Video-Based Benchmark Dataset for Fine-grained Personality Assessment with 15 Trait Dimensions},
  journal = {IEEE Transactions on Affective Computing},
  year    = {2026},
  doi     = {10.1109/TAFFC.2026.3698795}
}

Downloads last month: 33

Total file size:

1.04 GB