funterpie commited on
Commit
e4e108d
·
verified ·
1 Parent(s): 7dad56a

Update main.py

Browse files
Files changed (1) hide show
  1. main.py +21 -30
main.py CHANGED
@@ -1,60 +1,52 @@
1
  from fastapi import FastAPI, UploadFile, File, HTTPException
2
  from tensorflow.keras.models import load_model
 
 
3
  from PIL import Image
4
  import numpy as np
5
  import io
6
 
7
- # Initialize FastAPI app
8
- app = FastAPI(
9
- title="MNIST Digit Classifier API",
10
- version="1.0"
11
- )
12
 
13
- # Global model variable
14
  model = None
15
 
16
-
17
- # Load model at startup (SAFE WAY)
18
  @app.on_event("startup")
19
  def load_model_on_startup():
20
  global model
21
  try:
22
- model = load_model("mnist_model.h5")
23
- print("Model loaded successfully")
 
 
 
 
 
 
 
 
 
 
24
  except Exception as e:
25
- raise RuntimeError(f"Error loading model: {e}")
26
-
27
 
28
- # Health check
29
  @app.get("/")
30
  def read_root():
31
- return {
32
- "message": "MNIST Classification API is running. Send POST requests to /predict/"
33
- }
34
 
35
-
36
- # Prediction endpoint
37
  @app.post("/predict/")
38
  async def predict_digit(file: UploadFile = File(...)):
39
-
40
  if not file.content_type.startswith("image/"):
41
- raise HTTPException(status_code=400, detail="File provided is not an image.")
42
 
43
  try:
44
- # Read image
45
  contents = await file.read()
46
  image = Image.open(io.BytesIO(contents)).convert("L")
47
-
48
- # Resize to MNIST format
49
  image = image.resize((28, 28))
50
-
51
- # Convert to numpy
52
  image_array = np.array(image).astype("float32") / 255.0
53
-
54
- # Flatten for Dense model
55
  image_array = image_array.reshape(1, 784)
56
 
57
- # Predict
58
  prediction_probs = model.predict(image_array, verbose=0)
59
  predicted_class = int(np.argmax(prediction_probs))
60
  confidence = float(np.max(prediction_probs))
@@ -64,6 +56,5 @@ async def predict_digit(file: UploadFile = File(...)):
64
  "predicted_digit": predicted_class,
65
  "confidence": round(confidence, 4)
66
  }
67
-
68
  except Exception as e:
69
- raise HTTPException(status_code=500, detail=f"Error processing image: {str(e)}")
 
1
  from fastapi import FastAPI, UploadFile, File, HTTPException
2
  from tensorflow.keras.models import load_model
3
+ import tensorflow as tf
4
+ import os
5
  from PIL import Image
6
  import numpy as np
7
  import io
8
 
9
+ app = FastAPI(title="MNIST Digit Classifier API", version="1.0")
 
 
 
 
10
 
 
11
  model = None
12
 
 
 
13
  @app.on_event("startup")
14
  def load_model_on_startup():
15
  global model
16
  try:
17
+ # Try .keras first, fallback to .h5
18
+ model_path = "mnist_model.keras"
19
+ if not os.path.exists(model_path):
20
+ model_path = "mnist_model.h5"
21
+
22
+ print(f"Loading model from: {model_path}")
23
+ model = load_model(model_path, compile=False)
24
+ model.compile(optimizer='adam',
25
+ loss='sparse_categorical_crossentropy',
26
+ metrics=['accuracy'])
27
+ print("✅ Model loaded successfully!")
28
+ print(f"Input shape: {model.input_shape}")
29
  except Exception as e:
30
+ print(f"❌ Failed to load model: {e}")
31
+ raise RuntimeError(f"Model loading error: {e}")
32
 
 
33
  @app.get("/")
34
  def read_root():
35
+ return {"message": "MNIST API is running. POST to /predict/"}
 
 
36
 
 
 
37
  @app.post("/predict/")
38
  async def predict_digit(file: UploadFile = File(...)):
 
39
  if not file.content_type.startswith("image/"):
40
+ raise HTTPException(400, "File must be an image")
41
 
42
  try:
 
43
  contents = await file.read()
44
  image = Image.open(io.BytesIO(contents)).convert("L")
 
 
45
  image = image.resize((28, 28))
46
+
 
47
  image_array = np.array(image).astype("float32") / 255.0
 
 
48
  image_array = image_array.reshape(1, 784)
49
 
 
50
  prediction_probs = model.predict(image_array, verbose=0)
51
  predicted_class = int(np.argmax(prediction_probs))
52
  confidence = float(np.max(prediction_probs))
 
56
  "predicted_digit": predicted_class,
57
  "confidence": round(confidence, 4)
58
  }
 
59
  except Exception as e:
60
+ raise HTTPException(500, f"Processing error: {str(e)}")