Spaces:
Runtime error
Runtime error
Commit ·
b1899bd
0
Parent(s):
Initial commit
Browse files- Dockerfile.dockerfile +12 -0
- Procfile +1 -0
- README.md +8 -0
- app.py +38 -0
- requirements.txt +6 -0
- utils.py +14 -0
Dockerfile.dockerfile
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
FROM python:3.10
|
| 2 |
+
|
| 3 |
+
# Installer dépendances
|
| 4 |
+
WORKDIR /app
|
| 5 |
+
COPY requirements.txt .
|
| 6 |
+
RUN pip install --no-cache-dir -r requirements.txt
|
| 7 |
+
|
| 8 |
+
# Copier le code
|
| 9 |
+
COPY . .
|
| 10 |
+
|
| 11 |
+
# Lancer l'application avec Gunicorn
|
| 12 |
+
CMD ["gunicorn", "app:app", "--bind", "0.0.0.0:7860"]
|
Procfile
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
web: python app.py
|
README.md
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# API Alzheimer avec Flask + Hugging Face + Render
|
| 2 |
+
|
| 3 |
+
Cette API prend en entrée une image et retourne une prédiction en utilisant un modèle TensorFlow Lite stocké sur Hugging Face.
|
| 4 |
+
|
| 5 |
+
## Déploiement
|
| 6 |
+
- Code hébergé sur GitHub
|
| 7 |
+
- Modèle hébergé sur Hugging Face
|
| 8 |
+
- API déployée sur Render
|
app.py
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from flask import Flask, request, jsonify
|
| 2 |
+
import tensorflow as tf
|
| 3 |
+
import numpy as np
|
| 4 |
+
import os
|
| 5 |
+
from huggingface_hub import hf_hub_download
|
| 6 |
+
from utils import preprocess_image
|
| 7 |
+
|
| 8 |
+
app = Flask(__name__)
|
| 9 |
+
|
| 10 |
+
# Récupérer le modèle depuis Hugging Face Hub
|
| 11 |
+
MODEL_PATH = hf_hub_download(
|
| 12 |
+
repo_id="aroussya/alzheimer-tflite",
|
| 13 |
+
filename="alzheimer_model_float32.tflite",
|
| 14 |
+
token=os.getenv("HF_TOKEN") # le token est dans l'environnement
|
| 15 |
+
)
|
| 16 |
+
|
| 17 |
+
# Charger le modèle TFLite
|
| 18 |
+
interpreter = tf.lite.Interpreter(model_path=MODEL_PATH)
|
| 19 |
+
interpreter.allocate_tensors()
|
| 20 |
+
|
| 21 |
+
input_details = interpreter.get_input_details()
|
| 22 |
+
output_details = interpreter.get_output_details()
|
| 23 |
+
|
| 24 |
+
@app.route("/predict", methods=["POST"])
|
| 25 |
+
def predict():
|
| 26 |
+
data = request.get_json()
|
| 27 |
+
image = np.array(data['image'], dtype=np.float32)
|
| 28 |
+
image = preprocess_image(image)
|
| 29 |
+
|
| 30 |
+
interpreter.set_tensor(input_details[0]['index'], image)
|
| 31 |
+
interpreter.invoke()
|
| 32 |
+
output_data = interpreter.get_tensor(output_details[0]['index'])
|
| 33 |
+
prediction = output_data.tolist()[0]
|
| 34 |
+
|
| 35 |
+
return jsonify({"prediction": prediction})
|
| 36 |
+
|
| 37 |
+
if __name__ == "__main__":
|
| 38 |
+
app.run(host="0.0.0.0", port=7860)
|
requirements.txt
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
flask
|
| 2 |
+
numpy
|
| 3 |
+
tensorflow
|
| 4 |
+
huggingface-hub
|
| 5 |
+
gunicorn
|
| 6 |
+
opencv-python
|
utils.py
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import numpy as np
|
| 2 |
+
import cv2
|
| 3 |
+
|
| 4 |
+
def preprocess_image(image_array):
|
| 5 |
+
"""
|
| 6 |
+
Prépare l'image pour le modèle TFLite.
|
| 7 |
+
- Redimensionne à 128x128 (adapter si ton modèle a une autre taille d'entrée)
|
| 8 |
+
- Normalise entre 0 et 1
|
| 9 |
+
- Ajoute la dimension batch
|
| 10 |
+
"""
|
| 11 |
+
image = np.array(image_array, dtype=np.float32)
|
| 12 |
+
image = cv2.resize(image, (128, 128))
|
| 13 |
+
image = image / 255.0
|
| 14 |
+
return image.reshape(1, 128, 128, 3)
|