app.py

from fastai.vision.all import *
path = untar_data(URLs.PETS)/'images'
def is_cat(x): return x[0].isupper() 
dls = ImageDataLoaders.from_name_func('.',
    get_image_files(path), valid_pct=0.2, seed=42,
    label_func=is_cat,
    item_tfms=Resize(192))

learn = vision_learner(dls, resnet18, metrics=error_rate)
learn.fine_tune(3)

learn.export('model.pkl')

im = PILImage.create('dog.jpg')
im.thumbnail((192,192))
im

learn = load_learner('model.pkl')

learn.predict(im)

categories = ('Dog', 'Cat')

def classify_image(img):
    pred, idx, probs=learn.predict(img)
    return dict(zip(categories, map(float, probs)))

classify_image(im)

import gradio as gr

image = gr.inputs.Image(shape=(192,192))
label = gr.outputs.Label()
examples = ['dog.jpg', 'cat.jpeg', 'raccoon.jpg']

intf = gr.Interface(fn=classify_image, inputs=image, outputs=label,examples=examples)
intf.launch(inline=False)