G44mlops/ResNet-medmnist

ResNet model card (MedMNIST – OrganAMNIST)

Dataset

• —Collection: MedMNIST , task: OrganAMNIST (multi-class, 11 organs).
• —Image format: grayscale, 28×28 resized to 224×224 via MedMNIST_dataset.
• —Available samples: train 34,561; val 6,491; test 17,778.
• —Used for this run: random subset of 10,000 train images; full val set.

Training setup

• —Architecture: ResNet18
• —Epochs: 10
• —Batch size: 64
• —Optimizer: Adam (lr 0.001, weight_decay 0.0001)
• —Num workers: 4
• —Device: CUDA