MUFAC: Asian Facial Images for Age Classification

Over 63,000 Asian facial images annotated for age groups and personal identities, forming the MUFAC benchmark. The dataset was preprocessed to a 128x128 resolution and includes CSV files for structured training and evaluation. It was uploaded by Dasool and last updated on May 26, 2025.

Use Cases

• Training age classification models based on facial images.
• Benchmarking machine unlearning algorithms for facial recognition systems.
• Studying facial feature representation across different age groups.
• Evaluating model fairness and bias on a dataset of Asian faces.

Strengths

• Contains over 63,000 facial images.
• Images are preprocessed to a consistent 128x128 resolution.
• Includes annotations for both age groups and personal identities.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to its focus on Asian facial images.

Provenance

Source

Dasool on Hugging Face.

Collection Method

Likely aggregated and cleaned from existing sources for a machine unlearning benchmark.

Freshness

Last updated 2025-05-26 07:20:02.

Geography

Focus on Asian facial images.