app.py

__all__ = ['install', 'learn', 'categories', 'classify_image', 'classify_image_url', 'image', 'label', 'examples', 'intf', 'intf_text_url']



# Kept getting "No module named 'fastai'" from huggingface..workaround:
# https://stackoverflow.com/a/50255019
import subprocess
import sys

def install(package):
    subprocess.check_call([sys.executable, "-m", "pip", "install", package])

install("fastai")



from fastai.vision.all import *
#from fast.ai.vision.all import *
import gradio as gr

# def greet(name):
#    return "Hello " + name + "!!"

# iface = gr.Interface(fn=greet, inputs="text", outputs="text")
# iface.launch()




# from:
#  https://course.fast.ai/Lessons/lesson2.html
#  https://github.com/fastai/fastbook/blob/master/02_production.ipynb


#learn_inf = load_learner(path/'export.pkl')
#learn_inf.predict('images/grizzly.jpg')
#learn_inf.dls.vocab


# model created from: https://www.kaggle.com/code/zachwormgoor/stock-photo-recognizer
learn = load_learner('model - 712 images 3 epochs.pkl')

categories = ('amateur', 'stock')

#may have categories swapped?
#add probs to UI somewhere

def classify_image(img):
    tens = tensor(img) #fix apparently needed after fastai 2.7.11 released
    pred,idx,probs = learn.predict(tens)
    return dict(zip(categories, map(float,probs)))

from fastdownload import download_url
import os

def classify_image_url(url_text):
    try:
        dest = 'temp.jpg'
        download_url(url_text, dest, show_progress=False)
        im = Image.open(dest)
        img = im.to_thumb(256,256)
        #resize_images(dest, max_size=400, dest=dest)
        os.remove(dest)
        return classify_image(img)
    except:
        # in case there is any error, invalid URL or invalid image, etc., not sure how Gradio will handle a runtime exception so catching it to be safe
        return { categories[0]: 0.0, categories[1]: 0.0 }


def classify_image_url_debug(url_text):
    try:
        dest = 'temp.jpg'
        download_url(url_text, dest, show_progress=False)
        im = Image.open(dest)
        img = im.to_thumb(256,256)
        #resize_images(dest, max_size=400, dest=dest)
        os.remove(dest)
        temp = classify_image(img)
        return "Success: " + str(temp)
    except Exception as ex:
        # in case there is any error, invalid URL or invalid image, etc., not sure how Gradio will handle a runtime exception so catching it to be safe
        #return { categories[0]: 0.0, categories[1]: 0.0 }
        error = f"{type(ex).__name__} was raised: {ex}"
        return error;


    
    
image     = gr.inputs.Image(shape=(192, 192))
label     = gr.outputs.Label()
examples = ['stock.jpg', 'stock2.jpg', 'stock-easy.jpg', 'stock-easy2.jpg', 'combine-stock.jpg', 'washing-machine-stock.jpg', 'glasses-amateur.jpg', 'amateur.jpg', 'amateur2.jpg', 'amateur-easy.jpg', 'amateur-easy2.jpg', 'unsure.jpg']
# stock:       https://cdn.pixabay.com/photo/2022/02/11/12/12/pforphoto-7007231_1280.jpg
# stock2:      https://cdn.pixabay.com/photo/2019/09/25/13/57/business-4503747_960_720.jpg
# stock-easy:  https://scienceline.org/wp-content/uploads/2015/10/apple-926456_1280.jpg
# stock-easy2: https://ph-files.imgix.net/6b5cb0af-8bef-4700-b35f-8576e0ec58d8.jpeg?auto=format&fit=crop&frame=1&h=512&w=1024
# washing-machine-stock: https://live.staticflickr.com/7313/12775521983_e0a67aeb86_n.jpg
# combine-stock: https://live.staticflickr.com/5765/22059912468_ec5c315be4_b.jpg
# glasses-amateur: http://2.bp.blogspot.com/-dpygWXrSzRo/VfiIfzsWuxI/AAAAAAAAGvc/SNLnjV-IzO4/s1600/Green%2BTea%2Band%2BOrange%2BCollard%2BRecipe.jpg
# amateur:     https://cimg3.ibsrv.net/cimg/www.honda-tech.com/1024x768_85/479/AAApril-2-Control-Arms-01-58479.jpg
# amateur2:    https://www.iflash.xyz/images/diyMod/img7.jpg
# unsure:      https://en.wikipedia.org/wiki/Joshua_Bean


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


intf_text_url = gr.Interface(fn=classify_image_url, inputs="text", outputs=gr.outputs.Label())#, examples=example_urls)
intf_text_url_debug = gr.Interface(fn=classify_image_url_debug, inputs="text", outputs="text")
# intf_text_url.launch(inline=False)

# https://github.com/gradio-app/gradio/issues/450
gr.TabbedInterface( [intf, intf_text_url, intf_text_url_debug], ["Image file upload", "Image URL", "Image URL debug" ] ).launch()