Model weights and predictions for reproducible benchmarking experiments in MedMNIST v2

This data repository is associated with our GitHub code <code>weights_*.zip</code>: PyTorch, AutoKeras and Google AutoML Vision are provided for MedMNIST2D. PyTorch and AutoKeras are provided for MedMNIST3D. If you are using PyTorch model weights, please note that the ResNet18_224 / ResNet50_224 models are trained with images resized to 224 x 224 by <code>PIL.Image.NEAREST</code>. Snapshots for <code>auto-sklearn</code> are not uploaded due to the embarrassingly large model sizes (lots of model ensemble). <code>predictions.zip</code>: We also provide all standard prediction files by PyTorch, auto-sklearn, AutoKeras and Google AutoML Vision, which works with <code>medmnist.Evaluator</code>. Each file is named as <code>{flag}_{split}_[AUC]{auc:.3f}_[ACC]{acc:.3f}@{run}.csv</code>, e.g., <code>bloodmnist_test_[AUC]0.997_[ACC]0.957@autokeras_3.csv</code>.本数据集仓库与我们的GitHub代码包<code>weights_*.zip</code>配套：针对MedMNIST2D任务，我们提供了PyTorch、AutoKeras与Google AutoML Vision三类适配的模型权重；针对MedMNIST3D任务，则提供了PyTorch与AutoKeras两类适配的模型权重。 若您使用PyTorch模型权重，请注意：ResNet18_224与ResNet50_224模型均采用通过<code>PIL.Image.NEAREST</code>将图像缩放至224×224分辨率后的样本进行训练。 由于<code>auto-sklearn</code>模型体量异常庞大（包含大量集成模型），其快照文件未上传至本仓库。 <code>predictions.zip</code>：我们还提供了由PyTorch、auto-sklearn、AutoKeras及Google AutoML Vision生成的全部标准预测结果文件，该文件集可配合<code>medmnist.Evaluator</code>工具使用。所有文件均遵循如下命名规则：<code>{flag}_{split}_[AUC]{auc:.3f}_[ACC]{acc:.3f}@{run}.csv</code>，示例为<code>bloodmnist_test_[AUC]0.997_[ACC]0.957@autokeras_3.csv</code>。