Base model: https://huggingface.co/google/vit-base-patch16-224

This achieves around 86% on the testing set, you can use it as a baseline for further tuning.

The Stanford car dataset contains 16,185 images of 196 classes of cars. Classes are typically at the level of Make, Model, Year, e.g. 2012 Tesla Model S or 2012 BMW M3 coupe. The data is split into 8144 training images, 6,041 testing images, and 2000 validation images in this case.

** Please note: this dataset does not contain newer car models **

from transformers import AutoFeatureExtractor, AutoModelForImageClassification

extractor = AutoFeatureExtractor.from_pretrained("therealcyberlord/stanford-car-vit-patch16")
model = AutoModelForImageClassification.from_pretrained("therealcyberlord/stanford-car-vit-patch16")

3D Object Representations for Fine-Grained Categorization Jonathan Krause, Michael Stark, Jia Deng, Li Fei-Fei 4th IEEE Workshop on 3D Representation and Recognition, at ICCV 2013 (3dRR-13). Sydney, Australia. Dec. 8, 2013.