Update app.py
Browse files
app.py
CHANGED
|
@@ -1,11 +1,12 @@
|
|
| 1 |
from fastapi import FastAPI, UploadFile, File
|
|
|
|
| 2 |
from huggingface_hub import hf_hub_download
|
| 3 |
from keras.models import load_model
|
| 4 |
-
from tensorflow.keras.preprocessing import image
|
| 5 |
from tensorflow.keras.applications.inception_v3 import preprocess_input
|
| 6 |
import numpy as np
|
| 7 |
from PIL import Image
|
| 8 |
import io
|
|
|
|
| 9 |
|
| 10 |
app = FastAPI()
|
| 11 |
|
|
@@ -27,30 +28,58 @@ class_names = ['acanthoica', 'akashiwo', 'alexandrium', 'amoeba', 'amphidinium',
|
|
| 27 |
model_path = hf_hub_download(repo_id="Daniel00611/InceptionV3_72", filename="InceptionV3_72.keras")
|
| 28 |
model = load_model(model_path)
|
| 29 |
|
| 30 |
-
def preprocess_image(
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
img_array = np.array(img) # Convertir la imagen a un array de NumPy
|
| 36 |
-
img_array = np.expand_dims(img_array, axis=0) # A帽adir dimensi贸n extra para lote
|
| 37 |
return img_array
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
@app.post("/predict/")
|
| 40 |
async def predict(file: UploadFile = File(...)):
|
| 41 |
try:
|
| 42 |
-
|
| 43 |
-
|
|
|
|
| 44 |
|
| 45 |
-
# Realizar
|
| 46 |
predictions = model.predict(img_array)[0]
|
| 47 |
|
| 48 |
# Obtener el top 10 de predicciones
|
| 49 |
top_10_indices = predictions.argsort()[-10:][::-1]
|
| 50 |
top_10_classes = [class_names[i] for i in top_10_indices]
|
| 51 |
top_10_probabilities = predictions[top_10_indices]
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
|
| 53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
result = [{"class": top_10_classes[i], "probability": float(top_10_probabilities[i])} for i in range(10)]
|
| 55 |
return {"predictions": result}
|
| 56 |
|
|
|
|
| 1 |
from fastapi import FastAPI, UploadFile, File
|
| 2 |
+
from pydantic import BaseModel
|
| 3 |
from huggingface_hub import hf_hub_download
|
| 4 |
from keras.models import load_model
|
|
|
|
| 5 |
from tensorflow.keras.applications.inception_v3 import preprocess_input
|
| 6 |
import numpy as np
|
| 7 |
from PIL import Image
|
| 8 |
import io
|
| 9 |
+
import base64
|
| 10 |
|
| 11 |
app = FastAPI()
|
| 12 |
|
|
|
|
| 28 |
model_path = hf_hub_download(repo_id="Daniel00611/InceptionV3_72", filename="InceptionV3_72.keras")
|
| 29 |
model = load_model(model_path)
|
| 30 |
|
| 31 |
+
def preprocess_image(img, target_size=(299, 299)):
|
| 32 |
+
img = img.resize(target_size) # Redimensionar la imagen
|
| 33 |
+
img_array = np.array(img) # Convertir a array de NumPy
|
| 34 |
+
img_array = np.expand_dims(img_array, axis=0) # Expandir dimensi贸n para lote
|
| 35 |
+
img_array = preprocess_input(img_array) # Preprocesar la imagen
|
|
|
|
|
|
|
| 36 |
return img_array
|
| 37 |
+
|
| 38 |
+
# Modelo para recibir la imagen en formato Base64
|
| 39 |
+
class ImageBase64(BaseModel):
|
| 40 |
+
image_base64: str
|
| 41 |
+
|
| 42 |
+
# Ruta para im谩genes subidas como archivo
|
| 43 |
@app.post("/predict/")
|
| 44 |
async def predict(file: UploadFile = File(...)):
|
| 45 |
try:
|
| 46 |
+
# Leer la imagen subida
|
| 47 |
+
img = Image.open(io.BytesIO(await file.read()))
|
| 48 |
+
img_array = preprocess_image(img)
|
| 49 |
|
| 50 |
+
# Realizar predicci贸n
|
| 51 |
predictions = model.predict(img_array)[0]
|
| 52 |
|
| 53 |
# Obtener el top 10 de predicciones
|
| 54 |
top_10_indices = predictions.argsort()[-10:][::-1]
|
| 55 |
top_10_classes = [class_names[i] for i in top_10_indices]
|
| 56 |
top_10_probabilities = predictions[top_10_indices]
|
| 57 |
+
|
| 58 |
+
# Formar respuesta en formato JSON
|
| 59 |
+
result = [{"class": top_10_classes[i], "probability": float(top_10_probabilities[i])} for i in range(10)]
|
| 60 |
+
return {"predictions": result}
|
| 61 |
|
| 62 |
+
except Exception as e:
|
| 63 |
+
return {"error": str(e)}
|
| 64 |
+
|
| 65 |
+
# Ruta para im谩genes en formato Base64
|
| 66 |
+
@app.post("/predict_base64/")
|
| 67 |
+
async def predict_base64(image_data: ImageBase64):
|
| 68 |
+
try:
|
| 69 |
+
# Decodificar la imagen Base64
|
| 70 |
+
image_bytes = base64.b64decode(image_data.image_base64)
|
| 71 |
+
img = Image.open(io.BytesIO(image_bytes))
|
| 72 |
+
img_array = preprocess_image(img)
|
| 73 |
+
|
| 74 |
+
# Realizar predicci贸n
|
| 75 |
+
predictions = model.predict(img_array)[0]
|
| 76 |
+
|
| 77 |
+
# Obtener el top 10 de predicciones
|
| 78 |
+
top_10_indices = predictions.argsort()[-10:][::-1]
|
| 79 |
+
top_10_classes = [class_names[i] for i in top_10_indices]
|
| 80 |
+
top_10_probabilities = predictions[top_10_indices]
|
| 81 |
+
|
| 82 |
+
# Formar respuesta en formato JSON
|
| 83 |
result = [{"class": top_10_classes[i], "probability": float(top_10_probabilities[i])} for i in range(10)]
|
| 84 |
return {"predictions": result}
|
| 85 |
|