app.py

# %%
#|default_exp app

# %% [markdown]
# # Bear Classifier App
# 
# This notebook creates uses an exported model `export.pkl` for a bear classifier, to create a python script which can run the model on HuggingFace. 

# %%
#|export
from fastai.vision.all import *
import gradio as gr

# %% [markdown]
# Let's take a look at an example picture:

# %%
#im = PILImage.create('teddybear.jpg')
#im = PILImage.create('grizzly.jpg')
im = PILImage.create('blackbear.jpg')

im.thumbnail((192, 192))
im


# %% [markdown]
# Let's import the model and create the learner:

# %%
#|export
import pathlib
plt = platform.system()
if plt == 'Windows': pathlib.WindowsPath = pathlib.PosixPath
learn = load_learner('export.pkl')

# %% [markdown]
# With the learner we can to the predictions (inference):

# %%
learn.predict(im)

# %% [markdown]
# The available categories are contained in the vocab:

# %%
learn.dls.vocab

# %% [markdown]
# This is the function to classify the images:

# %%
#|export
def classify_image(img):
    pred,pred_idx,probs = learn.predict(img)
    return dict(zip(learn.dls.vocab, map(float, probs)))

# %% [markdown]
# Testing the function:

# %%
classify_image(im)

# %% [markdown]
# ## Gradio App
# 
# Now it is time to create the gradio app:

# %%
# commented, because it produced warnings

#image = gr.inputs.Image(shape=(192,192))
#label = gr.outputs.Label()

# %%
#|export
image = gr.components.Image(shape=(192,192))
label = gr.components.Label()
examples = ['teddybear.jpg', 'grizzly.jpg', 'blackbear.jpg']

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

# %%
intf.close()

# %% [markdown]
# ## Export
# 
# Finally, we export the code in the cells which are marked with `#|export`:

# %%


# %%
 

# %%