app.py

import numpy as np
import torch
from torchvision import transforms
from transformers import AutoModelForImageClassification
import gradio as gr
from PIL import Image

image_preprocess = transforms.Compose([
    transforms.Resize(224),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.5,0.5,0.5], std=[0.5,0.5,0.5])
    ])
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')


model_id = "PinChunPai/cat20_breed_fine_tune"
model = AutoModelForImageClassification.from_pretrained(model_id).to(device)

id2label=model.config.id2label
model.eval()
def classifier_app(image):
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image.astype('uint8'), 'RGB')


    image = image_preprocess(image)
    image = image.unsqueeze(0)
    image = image.to(device)


    with torch.no_grad():
      logits = model(image).logits

    prediction = torch.softmax(logits, dim=1)

    top_indices = torch.argsort(prediction, descending=True)[:].tolist()
    top1 = round( prediction[0][top_indices[0][0]].item(), 4)
    top2 = round( prediction[0][top_indices[0][1]].item(), 4)
    top3 = round( prediction[0][top_indices[0][2]].item(), 4)
    others = round(1.0 - top1 - top2-top3 , 4)

    len_1 = len(id2label[top_indices[0][0]])
    len_2 = len(id2label[top_indices[0][1]])
    len_3 = len(id2label[top_indices[0][2]])
    max_length = max(len_1, len_2, len_3, len('Others'))

    labels = [f'{id2label[top_indices[0][0]]}'.ljust(max_length), f'{id2label[top_indices[0][1]]}'.ljust(max_length), f'{id2label[top_indices[0][2]]}'.ljust(max_length), f'Others'.ljust(max_length)]
    preds = [top1, top2,top3 , others]
    result = {label : float(pred) for label, pred in zip(labels, preds)}

    return result
description =  f"This classifier assumes the picture is among the 20 breeds :{list(id2label.values())}"
article  = "First two examples are extracted from Kaggle dataset(https://www.kaggle.com/datasets/knucharat/pop-cats). Model is trained on the same dataset.  Project details can be found in https://github.com/PinChunPai/Cat_breeds_classification/tree/main"

interface = gr.Interface(
    fn=classifier_app,
    inputs="image",
    examples = ['cat_1.jpg','cat_2.jpg','cat_3.jpg'],
    outputs="label",
    title="Cat20_breeds",
    description=description,
    article = article 
)

interface.launch(share=True)