app.py

# Hide - Connects voila to current notebook
#!jupyter serverextension enable --sys-prefix voila

# Voila runs jpnb but hides the code cells and only displays the output (including the ipywidgets) as well as the markdown cells.

import streamlit as st
from fastai.vision.all import *
from fastai.vision.widgets import *

# Load the model
path = Path()
learn_inf = load_learner(path/'export.pkl')

# Title of the app
st.title('Forest Image Classifier')

# File uploader
uploaded_file = st.file_uploader("Select your forest image", type=['jpg', 'jpeg', 'png'])

# If an image has been uploaded
if uploaded_file is not None:
    # Display the uploaded image
    img = PILImage.create(uploaded_file)
    st.image(img.to_thumb(128, 128), caption='Uploaded Image', use_column_width=True)

    # Classify the image
    pred, pred_idx, probs = learn_inf.predict(img)
    
    # Display the prediction and probability
    st.write(f'Prediction: {pred}; Probability: {probs[pred_idx]:.04f}')


# 
# 
# import fastbook
# fastbook.setup_book()
# from fastbook import *
# from fastai.vision.widgets import *

# # Load the model
# path = Path()
# path.ls(file_exts='.pkl')

# learn_inf = load_learner(path/'export.pkl')

# # Create a button widget
# btn_upload = widgets.FileUpload()
# btn_upload

# # Create a clear_output widget
# out_pl = widgets.Output()
# out_pl.clear_output()

# # Create a label widget
# lbl_pred = widgets.Label()

# # Create a run button widget
# btn_run = widgets.Button(description='Classify')
# btn_run

# # Function to classify the image
# def on_click_classify(change):
#     img = PILImage.create(btn_upload.data[-1])
#     out_pl.clear_output()
#     with out_pl: display(img.to_thumb(128,128))
#     pred,pred_idx,probs = learn_inf.predict(img)
#     lbl_pred.value = f'Prediction: {pred}; Probability: {probs[pred_idx]:.04f}'

# btn_run.on_click(on_click_classify)

# # Create a Vertical Box (VBox) to hold the widgets
# btn_upload = widgets.FileUpload()
# VBox([widgets.Label('Select your forest!'),
#       btn_upload, btn_run, out_pl, lbl_pred])