Image classifier model

Download and Install the dependency

   !pip install -Uqq fastai
   !pip install -Uqq duckduckgo_search

!pip install -Uqq duckduckgo_search

Downloading Images with the DuckDuckGo Search API.

from duckduckgo_search import DDGS #
from fastcore.all import *

def search_images(keywords, max_images=200): return L(DDGS().images(keywords, max_results=max_images)).itemgot('image')

urls = search_images('biryani photos', max_images=1)
urls[0]

Download and display a sample biryani image

from fastdownload import download_url
dest = 'biryani.jpg'
download_url(urls[0], dest, show_progress=False)

from fastai.vision.all import *
im = Image.open(dest)
im.to_thumb(256,256)

Download and display the sample pulao image

download_url(search_images('pulao photos', max_images=1)[0], 'pulao.jpg', show_progress=False)
Image.open('pulao.jpg').to_thumb(256,256)

Download a few sample images and store them in a directory

searches = 'biryani','pulao'
path = Path('biryani_or_pulao')
from time import sleep

for o in searches:
    dest = (path/o)
    dest.mkdir(exist_ok=True, parents=True)
    download_images(dest, urls=search_images(f'{o} photo'))
    sleep(10)
    resize_images(path/o, max_size=400, dest=path/o)

Path('biryani_or_pulao').ls()

Path('biryani_or_pulao/biryani').ls()

Path('biryani_or_pulao/pulao').ls()

Using the DuckDuck API, we downloaded a total dataset of 365 food images, consisting of 174 biryani images and 191 pulao images.

Remove dead link images from the dataset.

failed = verify_images(get_image_files(path))
failed.map(Path.unlink)
len(failed)

Let's create and check our dataset of biryani and pulao images,

dls = DataBlock(
    blocks=(ImageBlock, CategoryBlock),
    get_items=get_image_files,
    description: "Image classifier using fastai"
    splitter=RandomSplitter(valid_pct=0.2, seed=42),
    get_y=parent_label,
    item_tfms=[Resize(192, method='squish')]
).dataloaders(path)

dls.show_batch(max_n=6)

Let's train it using the ResNet-18 architecture for 3 epochs

learn = vision_learner(dls, resnet18, metrics=error_rate)
learn.fine_tune(5)

Let's begin our predictions.

is_biryani,_,probs = learn.predict(PILImage.create('biryani.jpg'))
print(f"This is a: {is_biryani}.")
print(f"Probability it's a biryani: {probs[0]:.4f}\nProbability it's a pulao image: {probs[1]:.4f}")