PANDA vs. ImageNet Model Performance Metrics

A 5.5 KB Excel file containing comparative performance metrics for classification models. The dataset includes accuracies, Cohen's Kappa scores, and AUCs for models trained on the PANDA medical imaging dataset versus those trained on ImageNet. It was authored by Michail Georgios Papachristos and last updated in March 2026.

Use Cases

• Compare model accuracy scores between PANDA and ImageNet-trained models to assess domain-specific performance.
• Analyze Cohen's Kappa scores to evaluate agreement between model predictions and ground truth labels across different training datasets.
• Benchmark AUC values for models to understand their discriminative power on the target classification task.
• Use the provided metrics to inform the selection of a pre-training dataset (PANDA or ImageNet) for medical imaging applications.

Strengths

• Provides three distinct evaluation metrics (Accuracy, Cohen's Kappa, AUC) for direct comparison.
• Explicitly compares performance from two major training datasets (PANDA and ImageNet).
• Released under a permissive CC BY 4.0 license for broad reuse.

Limitations

• Extremely small scale at 5.5 KB, indicating a summary of results rather than raw model outputs or predictions.
• Contains only aggregated performance scores, lacking the underlying data, model architectures, or hyperparameters used.
• Unknown row count prevents assessment of the number of models or experiments compared.

Provenance

Source

Michail Georgios Papachristos via figshare.

Collection Method

Presumably derived from evaluating classification models trained on the PANDA and ImageNet datasets.

Freshness

Last updated March 2026.