HAM10000: Multi-Source Dermatoscopic Images of Pigmented Skin Lesions

HAM10000 is a large collection of multi-source dermatoscopic images of common pigmented skin lesions. The dataset was released to tackle the problem of small size and lack of diversity in available datasets for training neural networks for automated diagnosis. It was uploaded by author marmal88 to Hugging Face and last updated on 2023-01-25.

Use Cases

• Training neural networks for automated diagnosis based on dermatoscopic images.
• Benchmarking computer vision models for medical image classification of skin lesions.
• Addressing dataset size and diversity limitations in medical AI research.

Strengths

• Contains a large collection of dermatoscopic images.
• Explicitly addresses the problem of small and non-diverse datasets for training.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Original Paper and Dataset referenced, Kaggle dataset also referenced.

Collection Method

Multi-source collection of dermatoscopic images.

Freshness

Last updated 2023-01-25 02:21:28; freshness should be verified.