MicroViT-S2

ImageNet-1K pretrained MicroViT-S2 (10.0M params, 74.6% Top-1 accuracy).

Architecture: SHViT (Single-Head Vision Transformer) backbone. Source paper: https://arxiv.org/abs/2502.05800 Official repo: https://github.com/novendrastywn/MicroViT

Usage

from transformers import AutoModelForImageClassification, AutoImageProcessor

# Load ImageNet pretrained (1000 classes)
model = AutoModelForImageClassification.from_pretrained(
    "henriquequeirozcunha/microvit-s2",
    trust_remote_code=True,
)

# Fine-tune for binary classification
model = AutoModelForImageClassification.from_pretrained(
    "henriquequeirozcunha/microvit-s2",
    num_labels=2,
    ignore_mismatched_sizes=True,
    trust_remote_code=True,
)

Preprocessing: 224×224, ImageNet normalization (mean=[0.485,0.456,0.406], std=[0.229,0.224,0.225]).