import streamlit as st
import random
import pandas as pd
from PIL import Image
import requests
import json
from transformers import pipeline
import numpy as np
from transformers import AutoFeatureExtractor
from transformers import AutoModelForImageClassification
import plotly.graph_objects as go
import plotly
import re
st.set_page_config(layout='wide',
page_title='Food Category Classification & Recipes'
)
# Setting up Sidebar
sidebar_acc = ['App Description', 'About Project']
sidebar_acc_nav = st.sidebar.radio('**INFORMATION SECTION**', sidebar_acc)
if sidebar_acc_nav == 'App Description':
st.sidebar.markdown(" Food Category Classification Description
", unsafe_allow_html=True)
st.sidebar.markdown("This is a Food Category Image Classifier model that has been trained by [Kaludi](https://huggingface.co/Kaludi) to recognize **12** different categories of foods, which includes **Bread**, **Dairy**, **Dessert**, **Egg**, **Fried Food**, **Fruit**, **Meat**, **Noodles**, **Rice**, **Seafood**, **Soup**, and **Vegetable**. It can accurately classify an image of food into one of these categories by analyzing its visual features. This model can be used by food bloggers, restaurants, and recipe websites to quickly categorize and sort their food images, making it easier to manage their content and provide a better user experience.")
elif sidebar_acc_nav == 'About Project':
st.sidebar.markdown(" About Project
", unsafe_allow_html=True)
st.sidebar.markdown("
", unsafe_allow_html=True)
st.sidebar.markdown("Project Location:
", unsafe_allow_html=True)
st.sidebar.markdown("Model | Dataset
", unsafe_allow_html=True)
st.sidebar.markdown("
", unsafe_allow_html=True)
st.sidebar.markdown("Project Creators:
", unsafe_allow_html=True)
st.sidebar.markdown("AA
", unsafe_allow_html=True)
st.sidebar.markdown("AM
", unsafe_allow_html=True)
st.sidebar.markdown("BK
", unsafe_allow_html=True)
st.sidebar.markdown("DK
", unsafe_allow_html=True)
def main():
st.title("Food Category Classification & Recipes")
st.markdown("### Backgroud")
st.markdown("This is a Food Category Image Classifier model that has been trained by [Kaludi](https://huggingface.co/Kaludi) to recognize **12** different categories of foods, which includes **Bread**, **Dairy**, **Dessert**, **Egg**, **Fried Food**, **Fruit**, **Meat**, **Noodles**, **Rice**, **Seafood**, **Soup**, and **Vegetable**. It can accurately classify an image of food into one of these categories by analyzing its visual features. This model can be used by food bloggers, restaurants, and recipe websites to quickly categorize and sort their food images, making it easier to manage their content and provide a better user experience.")
st.header("Try it out!")
if st.checkbox("Show/Hide Examples"):
st.header("Example Images")
col1, col2, col3, col4 = st.columns(4)
with col1:
st.image("examples/example_0.jpg", width=260)
st.image("examples/example_1.jpg", width=260)
with col2:
st.image("examples/example_2.jpg", width=260)
st.image("examples/example_3.jpg", width=260)
with col3:
st.image("examples/example_4.jpg", width=260)
st.image("examples/example_5.jpg", width=260)
with col4:
st.image("examples/example_6.jpg", width=260)
st.image("examples/example_7.jpg", width=260)
# # display the text if the checkbox returns True value
# show_images = not show_images
# if show_images:
# st.header("Example Images")
# for image in images:
# st.image(image, width=260)
# select_health = st.radio("Select One (Not Functional Yet):", ["Regular", "Low-Calorie"], horizontal=True)
# Dropdown for Diet
diet_options = ['All', 'Gluten-Free', 'Vegan', 'Vegetarian', 'Dairy-Free']
diet = st.selectbox('Diet', diet_options)
# Dropdown for Cuisine
cuisine_options = ['All', 'African', 'Asian', 'Caribbean', 'Central American', 'Europe', 'Middle Eastern', 'North American', 'Oceanic', 'South American']
cuisine = st.selectbox('Cuisine', cuisine_options)
# Slider for Calories
calories = st.slider("Select Max Calories (Per Serving)", 25, 1000, 500)
# print the calories
st.write("Selected: **{}** Max Calories.".format(calories))
uploaded_file = st.file_uploader("Upload Files", type=['png','jpeg','jpg'])
loading_text = st.empty()
if uploaded_file != None:
loading_text.markdown("Loading...")
img = Image.open(uploaded_file)
extractor = AutoFeatureExtractor.from_pretrained("Kaludi/food-category-classification-v2.0")
model = AutoModelForImageClassification.from_pretrained("Kaludi/food-category-classification-v2.0")
inputs = extractor(img, return_tensors="pt")
outputs = model(**inputs)
# ...
loading_text.empty()
label_num=outputs.logits.softmax(1).argmax(1)
label_num=label_num.item()
probs = outputs.logits.softmax(dim=1)
percentage = round(probs[0, label_num].item() * 100, 2)
st.markdown("### Your Image:")
st.image(img, width=260)
st.write("The Predicted Classification is:")
if label_num==0:
st.write("**Bread** (" + f"{percentage}%)")
elif label_num==1:
st.write("**Dairy** (" + f"{percentage}%)")
elif label_num==2:
st.write("**Dessert** (" + f"{percentage}%)")
elif label_num==3:
st.write("**Egg** (" + f"{percentage}%)")
elif label_num==4:
st.write("**Fried Food** (" + f"{percentage}%)")
elif label_num==5:
st.write("**Fruit** (" + f"{percentage}%)")
elif label_num==6:
st.write("**Meat** (" + f"{percentage}%)")
elif label_num==7:
st.write("**Noodles** (" + f"{percentage}%)")
elif label_num==8:
st.write("**Rice** (" + f"{percentage}%)")
elif label_num==9:
st.write("**Seafood** (" + f"{percentage}%)")
elif label_num==10:
st.write("**Soup** (" + f"{percentage}%)")
else:
st.write("**Vegetable** (" + f"{percentage}%)")
st.write("You Selected **{}** For Diet and **{}** For Cuisine with Max".format(diet, cuisine), calories, "Calories For", ( "**Bread**" if label_num==0 else "**Dairy**" if label_num==1 else "**Dessert**" if label_num==2 else "**Egg**" if label_num==3 else "**Fried Food**" if label_num==4 else "**Fruit**" if label_num==5 else "**Meat**" if label_num==6 else "**Noodles**" if label_num==7 else "**Rice**" if label_num==8 else "**Seafood**" if label_num==9 else "**Soup**" if label_num==10 else "**Vegetable**"))
url = "https://alcksyjrmd.execute-api.us-east-2.amazonaws.com/default/nutrients_response"
category = ("Bread" if label_num==0 else "Dairy" if label_num==1 else "Dessert" if label_num==2 else "Egg" if label_num==3 else "Fried" if label_num==4 else "Fruit" if label_num==5 else "Meat" if label_num==6 else "Noodles" if label_num==7 else "Rice" if label_num==8 else "Seafood" if label_num==9 else "**Soup**" if label_num==10 else "Vegetable")
params = {"f": category, "k": str(calories)}
if diet != "All":
params["d"] = diet
if cuisine != "All":
params["c"] = cuisine
response = requests.get(url, params=params)
response_json = json.loads(response.content)
# Convert response_json to a list
response_json = list(response_json)
if len(response_json) == 0:
st.markdown("### No Recipe Found:")
st.write("**No recipes found. Please adjust your search criteria.**")
else:
st.markdown("### Recommended Recipe:")
if len(response_json) > 1:
random_recipe = random.choice(response_json)
nutrition_facts = random_recipe['Nutrition Facts']
st.write("**Title:** ", random_recipe['Title'])
if random_recipe['Image Link'].endswith(".jpg") or random_recipe['Image Link'].endswith(".jpeg") or random_recipe['Image Link'].endswith(".png"):
st.image(random_recipe['Image Link'], width=300)
else:
st.write("**Image Link:** ", random_recipe['Image Link'])
st.write("**Rating:** ", random_recipe['Rating'])
if random_recipe['Description'] != "Description not found":
st.write("**Description:** ", random_recipe['Description'])
st.write("**Ingredients:** ", random_recipe['Ingredients'])
st.write("**Recipe Facts:** ", random_recipe['Recipe Facts'])
st.write("**Directions:** ", random_recipe['Directions'])
nutrition_html = """
| Nutrition Facts (per serving) |
| Number of Servings: {servings} |
| Calories |
{calories} |
| Total Fat |
{total_fat} |
| Saturated Fat |
{saturated_fat} |
| Cholesterol |
{cholesterol} |
| Sodium |
{sodium} |
| Total Carbohydrate |
{total_carbohydrate} |
| Dietary Fiber |
{dietary_fiber} |
| Total Sugars |
{total_sugars} |
| Protein |
{protein} |
| Vitamin C |
{vitc} |
| Calcium |
{calc} |
| Iron |
{iron} |
| Potassium |
{pota} |
"""
# Use the nutrition HTML and format it with the values
formatted_html = nutrition_html.format(
calories=random_recipe['Calories'],
total_fat=random_recipe['Total Fat'],
saturated_fat=random_recipe['Saturated Fat'],
cholesterol=random_recipe['Cholesterol'],
sodium=random_recipe['Sodium'],
total_carbohydrate=random_recipe['Total Carbohydrate'],
dietary_fiber=random_recipe['Dietary Fiber'],
total_sugars=random_recipe['Total Sugars'],
servings=random_recipe['Number of Servings'],
vitc=random_recipe['Vitamin C'],
calc=random_recipe['Calcium'],
iron=random_recipe['Iron'],
pota=random_recipe['Potassium'],
protein=random_recipe['Protein']
)
# Define a function to apply the CSS styles to the table cells
def format_table(val):
return f"background-color: #133350; color: #fff; border: 1px solid #ddd; border-radius: .25rem; padding: .625rem .625rem 0; font-family: Helvetica; font-size: 1rem;"
with st.container():
# Add the nutrition table to the Streamlit app
st.write("Nutrition Facts
", unsafe_allow_html=True)
st.write(f"{formatted_html}
", unsafe_allow_html=True)
st.write("*The % Daily Value (DV) tells you how much a nutrient in a food serving contributes to a daily diet. 2,000 calories a day is used for general nutrition advice.
", unsafe_allow_html=True)
# extract only numeric values and convert mg to g
values = [
float(re.sub(r'[^\d.]+', '', random_recipe['Total Fat'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Saturated Fat'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Cholesterol'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Sodium'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Total Carbohydrate'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Dietary Fiber'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Total Sugars'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Protein'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Vitamin C'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Calcium'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Iron'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Potassium'])) / 1000
]
# create pie chart
labels = ['Total Fat', 'Saturated Fat', 'Cholesterol', 'Sodium', 'Total Carbohydrate', 'Dietary Fiber', 'Total Sugars', 'Protein', 'Vitamin C', 'Calcium', 'Iron', 'Potassium']
fig = go.Figure(data=[go.Pie(labels=labels, values=values)])
st.markdown("### Macronutrients Pie Chart ;) (In Grams)")
st.plotly_chart(fig)
st.write("**Tags:** ", random_recipe['Tags'])
st.write("**Recipe URL:** ", random_recipe['Recipe URLs'])
st.markdown("### JSON:")
st.write(response_json)
else:
st.markdown("### Recommended Recipe:")
st.write("**Title:** ", response_json[0]['Title'])
if response_json[0]['Image Link'].endswith(".jpg") or response_json[0]['Image Link'].endswith(".jpeg") or response_json[0]['Image Link'].endswith(".png"):
st.image(response_json[0]['Image Link'], width=300)
else:
st.write("**Image Link:** ", response_json[0]['Image Link'])
st.write("**Rating:** ", response_json[0]['Rating'])
if response_json[0]['Description'] != "Description not found":
st.write("**Description:** ", response_json[0]['Description'])
st.write("**Ingredients:** ", response_json[0]['Ingredients'])
st.write("**Recipe Facts:** ", response_json[0]['Recipe Facts'])
st.write("**Directions:** ", random_recipe['Directions'])
nutrition_html = """
| Nutrition Facts (per serving) |
| Number of Servings: {servings} |
| Calories |
{calories} |
| Total Fat |
{total_fat} |
| Saturated Fat |
{saturated_fat} |
| Cholesterol |
{cholesterol} |
| Sodium |
{sodium} |
| Total Carbohydrate |
{total_carbohydrate} |
| Dietary Fiber |
{dietary_fiber} |
| Total Sugars |
{total_sugars} |
| Protein |
{protein} |
| Vitamin C |
{vitc} |
| Calcium |
{calc} |
| Iron |
{iron} |
| Potassium |
{pota} |
"""
# Use the nutrition HTML and format it with the values
formatted_html = nutrition_html.format(
calories=random_recipe['Calories'],
total_fat=random_recipe['Total Fat'],
saturated_fat=random_recipe['Saturated Fat'],
cholesterol=random_recipe['Cholesterol'],
sodium=random_recipe['Sodium'],
total_carbohydrate=random_recipe['Total Carbohydrate'],
dietary_fiber=random_recipe['Dietary Fiber'],
total_sugars=random_recipe['Total Sugars'],
servings=random_recipe['Number of Servings'],
vitc=random_recipe['Vitamin C'],
calc=random_recipe['Calcium'],
iron=random_recipe['Iron'],
pota=random_recipe['Potassium'],
protein=random_recipe['Protein']
)
# Define a function to apply the CSS styles to the table cells
def format_table(val):
return f"background-color: #133350; color: #fff; border: 1px solid #ddd; border-radius: .25rem; padding: .625rem .625rem 0; font-family: Helvetica; font-size: 1rem;"
with st.container():
# Add the nutrition table to the Streamlit app
st.write("Nutrition Facts
", unsafe_allow_html=True)
st.write(f"{formatted_html}
", unsafe_allow_html=True)
st.write("*The % Daily Value (DV) tells you how much a nutrient in a food serving contributes to a daily diet. 2,000 calories a day is used for general nutrition advice.
", unsafe_allow_html=True)
# extract only numeric values and convert mg to g
values = [
float(re.sub(r'[^\d.]+', '', random_recipe['Total Fat'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Saturated Fat'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Cholesterol'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Sodium'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Total Carbohydrate'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Dietary Fiber'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Total Sugars'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Protein'])),
float(re.sub(r'[^\d.]+', '', random_recipe['Vitamin C'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Calcium'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Iron'])) / 1000,
float(re.sub(r'[^\d.]+', '', random_recipe['Potassium'])) / 1000
]
# create pie chart
labels = ['Total Fat', 'Saturated Fat', 'Cholesterol', 'Sodium', 'Total Carbohydrate', 'Dietary Fiber', 'Total Sugars', 'Protein', 'Vitamin C', 'Calcium', 'Iron', 'Potassium']
fig = go.Figure(data=[go.Pie(labels=labels, values=values)])
st.markdown("### Macronutrients Pie Chart ;) (In Grams)")
st.plotly_chart(fig)
st.write("**Tags:** ", random_recipe['Tags'])
st.write("**Recipe URL:** ", random_recipe['Recipe URLs'])
st.markdown("### JSON:")
st.write(response_json)
st.markdown("
", unsafe_allow_html=True)
st.markdown("Github | HuggingFace
", unsafe_allow_html=True)
if __name__ == '__main__':
main()
## Fetch random recipes:
## Fetch the healthy recipe
#response = requests.get("test.org/test.json")
#data = response.json()
#healthy_recipes = [recipe for recipe in data['recipe'] if recipe['calories'] < int(calories)]
#healthy_recipe = random.choice(healthy_recipes)
## Fetch the unhealthy recipe
#unhealthy_recipes = [recipe for recipe in data['recipe'] if recipe['calories'] < int(calories)]
#unhealthy_recipe = random.choice(unhealthy_recipes)
## Add code to display the healthy and unhealthy recipes
#st.write("Healthy Recipe: ", healthy_recipe['name'], "Calories: ", healthy_recipe['calories'])
#st.write("Unhealthy Recipe: ", unhealthy_recipe['name'], "Calories: ", unhealthy_recipe['calories'])