Fine-tuning 12 ImageNet- or language-pretrained vision models on MedMNIST shows ViT-based models generally beat CNNs, and end-to-end fine-tuning beats linear probing, but the training pipeline changes the ranking.
On biases in a UK B iobank-based retinal image classification model
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Rethinking Foundation Models for Medical Image Classification through a Benchmark Study on MedMNIST
Fine-tuning 12 ImageNet- or language-pretrained vision models on MedMNIST shows ViT-based models generally beat CNNs, and end-to-end fine-tuning beats linear probing, but the training pipeline changes the ranking.