app.py

import requests  # type: ignore
import streamlit as st  # type: ignore
from PIL import Image, ImageDraw, ImageFont, ImageOps  # type: ignore
import io
import os
import random

API_URL = "https://labs.dev.bigml.io/andromeda/"
API_USERNAME = os.getenv("BIGML_USERNAME")
API_KEY = os.getenv("BIGML_API_KEY")
API_AUTH = f"username={API_USERNAME};api_key={API_KEY}"
FONT = ImageFont.truetype("img/roboto.ttf", 25)
MODEL = "deepnet/5JidvaoVsQ28fCiJg3tBge21vwS"
PREDICTION_THRESHOLD = 0.4
HEALTHY_CLASSES =  ["Blueberry leaf", "Peach leaf", "Raspberry leaf", "Strawberry leaf",
                    "Tomato leaf", "Bell_pepper leaf"]
DISEASE_CLASSES = ["Tomato leaf yellow virus", "Tomato Septoria leaf spot",
                   "Corn leaf blight", "Potato leaf early blight"]


def resize(img, width):
    """ Resize an imge to a given width maintaining aspect ratio """
    percent = width / float(img.size[0])
    return img.resize((width, int((float(img.size[1]) * float(percent)))))


API_URL = "https://labs.dev.bigml.io/andromeda/"
API_USERNAME = os.getenv("BIGML_USERNAME")
API_KEY = os.getenv("BIGML_API_KEY")
API_AUTH = f"username={API_USERNAME};api_key={API_KEY}"

def detection(uploaded_file):
    # Upload image to BigML as a source
    source_response = requests.post(
        f"{API_URL}source?{API_AUTH}",
        files={"file": ("plant_image", uploaded_file)}
    )
    source = source_response.json()["resource"]
    # Generate prediction data
    data = {"model": MODEL, "input_data": {"000002": source}}
    response = requests.post(f"{API_URL}prediction?{API_AUTH}", json=data)
    regions = response.json()["prediction"].get("000000", [])
    # Remove the source, we don't need it any more
    requests.delete(f"{API_URL}{source}?{API_AUTH}")
    return [r for r in regions if r[5]>PREDICTION_THRESHOLD]


def draw_predictions(pil_image, boxes):
    """ Draw BigML predictions in the image, adding a black border too """
    w, h = pil_image.size
    draw = ImageDraw.Draw(pil_image)
    for box in boxes:
        label, xmin,ymin, xmax, ymax, confidence = box
        draw.rectangle(((xmin*w, ymin*h), (xmax*w, ymax*h)), width=9, outline="#eee")
        draw.text(
            (xmin*w+20, ymin*h+random.randint(10, 40)),
            f"{label}: {str(confidence)[:3]}", font=FONT,  fill="#eee"
        )
    return ImageOps.expand(pil_image ,border=50,fill='black')


def gen_message(boxes):
    """ Generate output message for predictions """
    labels = set([box[0] for box in boxes])
    healthy = labels.intersection(set(HEALTHY_CLASSES))
    diseases = labels.intersection(set(DISEASE_CLASSES))
    if len(diseases) > 0:        
        st.warning(f"🦠 Your plants needs a doctor! Found **{','.join(diseases)}**!")
    elif len(healthy) > 0:
        st.success(f"🪴 Your plants have good health! Found **{','.join(healthy)}**!")
    else:
        st.error("No plant was found")


st.set_page_config(
    layout="wide",
    page_title="Plant Disease Detection",
    page_icon="🌱",
)

# Sidebar information
description = """ Detect leafs from different kinds of plants and
diagnose common **diseases** in them.  """
image = Image.open('img/BigML.png')
st.sidebar.image(image, width=100)
st.sidebar.write(description)
st.sidebar.write("Powered by [BigML](https://bigml.com)")

# Page title
st.title("🌱 BigML Plant Disease Detection")
classes = "HEALTHY:\n"
for leaf in HEALTHY_CLASSES:
    classes += f"- {leaf}\n"
classes += "\nDISEASES:\n"
for leaf in DISEASE_CLASSES:
    classes += f"- {leaf}\n"

with st.expander("⚠️ Disease detection model was trained with a small dataset. It can be inaccurate sometimes. It should be able to find the following classes: "):
    st.write(classes) 


left, right = st.columns(2)

# Example images
examples = {
    "Example Tomato": "img/tomato-septoria-3-500x384.jpg",
    "Example Raspberry": "img/plant8.jpg",
    "Example Potato": "img/Potato-Blight-on-Leaves.jpg",    
    "Example Bell pepper": "img/Bell-pepper-leaves.jpg"
}

with left.expander(label="Example Plants", expanded=True):
    option = st.selectbox('Choose one example image...', examples.keys(),index=0)
    clicked = st.button("Diagnose selected image")
    if clicked:
        example_file = open(examples[option], 'rb')

# File uploader
msg = "Or upload your plant image..."
with right.form("submit", clear_on_submit=True):
    uploaded_file = st.file_uploader(msg, type=["png ", "jpg", "jpeg"])
    submitted = st.form_submit_button("Diagnose uploaded image")


file_to_predict = None
if clicked and example_file:
    file_to_predict = example_file
elif uploaded_file and submitted:
    file_to_predict = uploaded_file


# Prediction Output
if file_to_predict:
    st.subheader("Detection result")
    with st.spinner('Diagnose in progress. Please wait...'):
        boxes = detection(file_to_predict)
        image = resize(Image.open(file_to_predict), 1000)
        output_image = draw_predictions(image, boxes)
        gen_message(boxes)
        st.image(output_image, width=700)
        uploaded_file = None