Eyepacs+APTOS+IDRiD+DDR: Retinal Images Resized for EfficientNet

Eyepacs, APTOS, IDRiD, and DDR datasets have been resized using a Letterbox technique. The collection combines multiple sources of retinal fundus images, likely for diabetic retinopathy detection. The dataset is hosted on Kaggle, but specifics on size, license, and authorship are unknown.

Use Cases

• Train diabetic retinopathy classification models based on the combined retinal image data.
• Benchmark image preprocessing techniques, specifically letterbox resizing, for medical computer vision tasks.
• Develop transfer learning models using the EfficientNet architecture on preprocessed medical imagery.

Strengths

• Combines four established retinal image datasets (Eyepacs, APTOS, IDRiD, DDR), potentially increasing sample diversity.
• Images have been preprocessed with a specific technique (Letterbox resizing) suitable for the EfficientNet architecture.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• License and authorship details are unknown, which may restrict usage rights.

Provenance

Source

Kaggle

Collection Method

Resizing and combination of existing public datasets (Eyepacs, APTOS, IDRiD, DDR).