ODIR: Ocular Disease Intelligent Recognition Database for 5,000 Patients

5,000 patient records form the Ocular Disease Intelligent Recognition (ODIR) database, containing color fundus photographs from both eyes and doctors' diagnostic keywords. The dataset was uploaded to Hugging Face by user 'bumbledeep' and was last updated in April 2025. All images originate from a Kaggle dataset, with the contributor noting some minor modifications to the metadata.

Use Cases

• Training image classification models for ocular diseases based on color fundus photographs.
• Developing multimodal diagnostic systems that combine image data with textual diagnostic keywords.
• Conducting clinical research on patient demographics and disease co-occurrence mentioned in the description.

Strengths

• Contains data from 5,000 patients, providing a substantial sample size for analysis.
• Includes paired image data (left and right eye fundus photographs) and associated diagnostic text for each patient.

Limitations

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

Provenance

Source

Kaggle (original source), Hugging Face (hosted by user 'bumbledeep').

Collection Method

Structured ophthalmic database of patient records, with images and diagnostic keywords from doctors.

Freshness

Last updated 2025-04-08 12:07:10; freshness should be verified.