ILYAS72066/Chunky_Panday

1from fastapi import FastAPI, UploadFile, File
2import numpy as np
3from PIL import Image
4import io
5from tensorflow import keras
6
7# Load model
8model = keras.models.load_model('model.h5')
9
10# Class labels (ordered list)
11class_labels = [
12    'Apple__Apple_scab', 'Apple_Black_rot', 'Apple_Cedar_apple_rust', 'Apple__healthy',
13    'Blueberry__healthy', 'Cherry(including_sour)Powdery_mildew', 'Cherry(including_sour)_healthy',
14    'Corn_(maize)Cercospora_leaf_spot Gray_leaf_spot', 'Corn(maize)Common_rust',
15    'Corn_(maize)Northern_Leaf_Blight', 'Corn(maize)healthy', 'Grape__Black_rot',
16    'Grape__Esca(Black_Measles)', 'Grape__Leaf_blight(Isariopsis_Leaf_Spot)', 'Grape___healthy',
17    'Orange__Haunglongbing(Citrus_greening)', 'Peach__Bacterial_spot', 'Peach__healthy',
18    'Pepper,bell_Bacterial_spot', 'Pepper,_bell_healthy', 'Potato_Early_blight', 'Potato__Late_blight',
19    'Potato__healthy', 'Raspberry_healthy', 'Soybean_healthy', 'Squash__Powdery_mildew',
20    'Strawberry__Leaf_scorch', 'Strawberry_healthy', 'Tomato_Bacterial_spot', 'Tomato__Early_blight',
21    'Tomato__Late_blight', 'Tomato_Leaf_Mold', 'Tomato__Septoria_leaf_spot',
22    'Tomato__Spider_mites Two-spotted_spider_mite', 'Tomato__Target_Spot',
23    'Tomato__Tomato_Yellow_Leaf_Curl_Virus', 'Tomato_Tomato_mosaic_virus', 'Tomato__healthy'
24]
25
26# Initialize FastAPI app
27app = FastAPI()
28
29# Preprocess function
30def preprocess_image(image_bytes):
31    image = Image.open(io.BytesIO(image_bytes)).convert('RGB')
32    image = image.resize((224, 224))
33    img_array = np.array(image) / 255.0  # normalize (assuming your model was trained with normalization)
34    img_array = np.expand_dims(img_array, axis=0)  # add batch dimension
35    return img_array
36
37@app.post("/predict")
38async def predict(file: UploadFile = File(...)):
39    image_bytes = await file.read()
40    img_array = preprocess_image(image_bytes)
41    predictions = model.predict(img_array)
42    predicted_class = class_labels[np.argmax(predictions)]
43    confidence = float(np.max(predictions))
44    return {"prediction": predicted_class, "confidence": confidence}
45