Misupatel/vit-imagenette

ViT-Base fine-tuned on Imagenette

Fine-tuned from google/vit-base-patch16-224 on the Imagenette 160 px dataset.

Metric	Value
Val accuracy	99.52%
Val loss	0.017721
Best epoch	9
Classes	10

Classes

tench, English springer, cassette player, chain saw, church, French horn, garbage truck, gas pump, golf ball, parachute

Usage

from transformers import AutoModelForImageClassification, ViTImageProcessor
from PIL import Image
import torch

model     = AutoModelForImageClassification.from_pretrained("Misupatel/vit-imagenette")
processor = ViTImageProcessor.from_pretrained("Misupatel/vit-imagenette")
model.eval()

image  = Image.open("your_image.jpg").convert("RGB")
inputs = processor(images=image, return_tensors="pt")

with torch.no_grad():
    logits = model(**inputs).logits

predicted_class = model.config.id2label[logits.argmax(-1).item()]
print(predicted_class)

Training details

• Optimizer: AdamW, lr=2e-5
• Epochs: up to 10 (early stopping, patience=3)
• Batch size: 32
• Augmentation: RandomResizedCrop(224), RandomHorizontalFlip, ColorJitter, RandomRotation(15°)
• Normalisation: mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]