Spaces:
Running
Running
Create main.py
Browse files
main.py
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import numpy as np
|
3 |
+
import tensorflow as tf
|
4 |
+
from fastapi import FastAPI, File, UploadFile
|
5 |
+
from fastapi.responses import JSONResponse
|
6 |
+
from io import BytesIO
|
7 |
+
from PIL import Image
|
8 |
+
from tensorflow.keras.preprocessing.image import img_to_array
|
9 |
+
from tensorflow.keras.applications import resnet50
|
10 |
+
from tensorflow.keras.applications.resnet50 import preprocess_input
|
11 |
+
import uvicorn
|
12 |
+
|
13 |
+
# Initialize FastAPI app
|
14 |
+
app = FastAPI()
|
15 |
+
|
16 |
+
# Model and class information
|
17 |
+
model_path = "model.keras"
|
18 |
+
class_indices = {0: 'glaucoma', 1: 'normal'}
|
19 |
+
|
20 |
+
# Load the model if it exists
|
21 |
+
if os.path.exists(model_path):
|
22 |
+
model = tf.keras.models.load_model(model_path)
|
23 |
+
print("Model loaded successfully.")
|
24 |
+
else:
|
25 |
+
print(f"Model file not found at {model_path}. Please upload the model.")
|
26 |
+
|
27 |
+
# Function to predict glaucoma in an image and return the class name
|
28 |
+
def predict_image(image_data):
|
29 |
+
try:
|
30 |
+
# Load the image from binary data
|
31 |
+
img = Image.open(BytesIO(image_data))
|
32 |
+
# Resize the image to the target size
|
33 |
+
img = img.resize((224, 224))
|
34 |
+
# Convert image to array format for the model
|
35 |
+
img_array = img_to_array(img)
|
36 |
+
img_array = np.expand_dims(img_array, axis=0)
|
37 |
+
img_array = preprocess_input(img_array)
|
38 |
+
|
39 |
+
# Make prediction
|
40 |
+
prediction = model.predict(img_array)
|
41 |
+
predicted_class = np.argmax(prediction[0])
|
42 |
+
class_name = class_indices[predicted_class] # Map to class name
|
43 |
+
return class_name
|
44 |
+
except Exception as e:
|
45 |
+
print("Prediction error:", e)
|
46 |
+
return "Error during prediction"
|
47 |
+
|
48 |
+
# Route for health check
|
49 |
+
@app.get("/health")
|
50 |
+
async def api_health_check():
|
51 |
+
return JSONResponse(content={"status": "Service is running"})
|
52 |
+
|
53 |
+
# Route for prediction using image via API
|
54 |
+
@app.post("/predict")
|
55 |
+
async def api_predict_image(file: UploadFile = File(...)):
|
56 |
+
try:
|
57 |
+
# Read the image file as binary data
|
58 |
+
image_data = await file.read()
|
59 |
+
|
60 |
+
# Call the prediction function with the image data
|
61 |
+
prediction = predict_image(image_data)
|
62 |
+
|
63 |
+
return JSONResponse(content={"prediction": prediction})
|
64 |
+
except Exception as e:
|
65 |
+
return JSONResponse(content={"error": str(e)})
|
66 |
+
|
67 |
+
# Run the FastAPI app
|
68 |
+
if __name__ == "__main__":
|
69 |
+
uvicorn.run(app, host="0.0.0.0", port=7860)
|