Datasets:

marcelohaps
/

agedb

license: other
task_categories:
  - image-classification
  - image-to-image
tags:
  - face-recognition
  - age-estimation
  - agedb
  - imagefolder
pretty_name: AgeDB HF-ready
size_categories:
  - 10K<n<100K

AgeDB HF-ready

This folder packages the local AgeDB images as a Hugging Face imagefolder dataset.

Dataset Structure

train/images/<shard>/: AgeDB image files split into shard directories.
train/metadata.csv: per-image labels and metadata.

The labels are derived from the AgeDB filename pattern:

<image_id>_<identity>_<age>_<gender>.jpg

Columns

file_name: relative image path used by Hugging Face ImageFolder, such as images/000/example.jpg.
label: numeric identity label.
label_name: identity name corresponding to label.
identity: normalized identity name.
image_id: numeric id from the original filename.
age: age annotation.
age_decade: decade bucket, such as 20s.
age_group: broad age bucket: child, teen, young_adult, adult, or senior.
gender: original compact gender label, f or m.
gender_label: expanded gender label.
source_filename: original AgeDB filename.

Local Stats

Images: 16488
Identities: 567
Age range: 1-101
Female images: 6700
Male images: 9788

Loading

from datasets import load_dataset

dataset = load_dataset("imagefolder", data_dir="data/evaluation/huggingface/agedb")
train = dataset["train"]

Notes

AgeDB is described by its authors as an in-the-wild face dataset annotated with identity, age, and gender. Check the original dataset terms before publishing or redistributing it.