app.py

import os
import streamlit as st
from PIL import Image

from inference import get_predictions, get_nearest_k


st.title('Fashion accessories prediction and search Demo')

tot_index = 44065
# multiple folders to handle 10,000 files folder limit in git
sample_path1 = './data/small_images_0_9999'
sample_path2 = './data/small_images_10000_19999'
sample_path3 = './data/small_images_20000_29999'
sample_path4 = './data/small_images_30000_39999'
sample_path5 = './data/small_images_40000_49999'

if 'image_index' not in st.session_state:
    st.session_state['image_index'] = 0

if 'which_button' not in st.session_state:
    st.session_state['which_button'] = 'sample_button'

sample_col, upload_col = st.tabs(['Select from sample images', 'Upload file'])
with upload_col:
    use_uploaded_image = True
    uploaded_file = st.file_uploader("Select a picture from your computer(png/jpg) :", type=['png', 'jpg', 'jpeg'])
    if uploaded_file is not None:
        img = Image.open(uploaded_file)
        st.image(img, caption='Uploaded Image')
        use_uploaded_image = st.button("Use uploaded image")
        if use_uploaded_image is True:
            st.session_state['which_button'] = 'upload_button'

with sample_col:
    use_sample_image = True
    st.write("Select one from these available samples: ")
    current_index = st.session_state['image_index']

    prev_button, next_button = st.columns(2)
    with prev_button:
        prev = st.button('prev_image')
    with next_button:
        next = st.button('next_image')
    if prev:
        current_index = (current_index - 1) % tot_index
    if next:
        current_index = (current_index + 1) % tot_index
    st.session_state['image_index'] = current_index
    if current_index < 9999:
        sample_path = sample_path1
    elif current_index < 19998:
        sample_path = sample_path2
    elif current_index < 29997:
        sample_path = sample_path3
    elif current_index < 39996:
        sample_path = sample_path4
    else:
        sample_path = sample_path5
    sample_image = Image.open(os.path.join(sample_path, str(current_index)+'.jpg'))
    st.image(sample_image, caption='Chosen image')

    use_sample_image = st.button("Use this Sample")
    if use_sample_image is True:
        st.session_state['which_button'] = 'sample_button'

classification_button, search_button = st.columns(2)
with classification_button:
    predict_clicked = st.button("Get categories predictions")

with search_button:
    search_clicked = st.button("Get similar looking products")

if predict_clicked:
    which_button = st.session_state['which_button']
    if which_button == 'sample_button':
        predictions = get_predictions(sample_image)
    elif which_button == 'upload_button':
        predictions = get_predictions(img)
    st.markdown('**The model predictions along with their probabilities are :**')
    st.table(predictions)
elif search_clicked:
    which_button = st.session_state['which_button']
    if which_button == 'sample_button':
        top_k_preds = get_nearest_k(sample_image)
    elif which_button == 'upload_button':
        top_k_preds = get_nearest_k(img)
    all_distances = top_k_preds[0][0]
    all_valid_distances = [dist for dist in all_distances if dist < 300]
    pred_to_show = len(all_valid_distances)
    st.markdown('**The top 5 similar product predictions are :**')
    if pred_to_show == 0:
        st.markdown('No similar visually looking similar products found in the database.')
    else:
        pred_cols = st.columns(pred_to_show)
        for i in range(len(pred_cols)):
            with pred_cols[i]:
                nearest_index = top_k_preds[1][0][i]
                if nearest_index < 9999:
                    the_path = sample_path1
                elif nearest_index < 19998:
                    the_path = sample_path2
                elif nearest_index < 29997:
                    the_path = sample_path3
                elif nearest_index < 39996:
                    the_path = sample_path4
                else:
                    the_path = sample_path5
                temp_img = Image.open(os.path.join(the_path, str(nearest_index)+'.jpg'))
                st.image(temp_img, caption=str(round(top_k_preds[0][0][i], 2))+' distance')