Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import tensorflow as tf
|
| 2 |
+
import numpy as np
|
| 3 |
+
import gradio as gr
|
| 4 |
+
from tensorflow.keras.preprocessing import image
|
| 5 |
+
from sklearn.preprocessing import LabelEncoder
|
| 6 |
+
import matplotlib.pyplot as plt
|
| 7 |
+
from PIL import Image
|
| 8 |
+
|
| 9 |
+
# Dictionnaire des sous-catégories
|
| 10 |
+
subcategory_dict = {
|
| 11 |
+
"Furniture": ["Home Decor"],
|
| 12 |
+
"Home Decor": [
|
| 13 |
+
"Poufs & Ottomans", "Rugs", "Antique items", "Brass Lamps",
|
| 14 |
+
"Candle Holders", "Pottery", "Kilim poufs", "Pillow Covers",
|
| 15 |
+
"Wall Decor", "Straw Lamps"
|
| 16 |
+
],
|
| 17 |
+
# Ajoutez d'autres catégories ici
|
| 18 |
+
}
|
| 19 |
+
|
| 20 |
+
# Fonction pour charger et prétraiter l'image
|
| 21 |
+
def preprocess_image(img):
|
| 22 |
+
img = img.resize((224, 224)) # Redimensionner
|
| 23 |
+
img_array = np.array(img) / 255.0 # Normaliser
|
| 24 |
+
img_array = np.expand_dims(img_array, axis=0) # Ajouter une dimension batch
|
| 25 |
+
return img_array
|
| 26 |
+
|
| 27 |
+
# Fonction pour prédire la catégorie, le prix et la sous-catégorie
|
| 28 |
+
def predict_image(img):
|
| 29 |
+
# Prétraiter l'image
|
| 30 |
+
img_array = preprocess_image(img)
|
| 31 |
+
|
| 32 |
+
# Faire les prédictions
|
| 33 |
+
category_pred, price_pred = model.predict(img_array)
|
| 34 |
+
|
| 35 |
+
# Décoder la catégorie
|
| 36 |
+
category_pred_class = np.argmax(category_pred, axis=1)[0]
|
| 37 |
+
category_name = label_encoder.inverse_transform([category_pred_class])[0]
|
| 38 |
+
|
| 39 |
+
# Trouver les sous-catégories correspondantes
|
| 40 |
+
subcategories = subcategory_dict.get(category_name, [])
|
| 41 |
+
|
| 42 |
+
# Préparer les résultats
|
| 43 |
+
results = {
|
| 44 |
+
"Category": category_name,
|
| 45 |
+
"Price ($)": f"{price_pred[0][0]:.2f}",
|
| 46 |
+
"Subcategories": subcategories
|
| 47 |
+
}
|
| 48 |
+
return results
|
| 49 |
+
|
| 50 |
+
# Charger le modèle pré-entraîné
|
| 51 |
+
# Assurez-vous que le chemin du modèle et de l'encodeur sont corrects
|
| 52 |
+
model = tf.keras.models.load_model('trained_model.h5')
|
| 53 |
+
label_encoder = LabelEncoder()
|
| 54 |
+
label_encoder.classes_ = np.load('path_to_label_encoder_classes.npy')
|
| 55 |
+
|
| 56 |
+
# Interface Gradio
|
| 57 |
+
interface = gr.Interface(
|
| 58 |
+
fn=predict_image,
|
| 59 |
+
inputs=gr.Image(type="pil"),
|
| 60 |
+
outputs=[
|
| 61 |
+
gr.Label(label="Category"),
|
| 62 |
+
gr.Text(label="Price ($)"),
|
| 63 |
+
gr.Text(label="Subcategories")
|
| 64 |
+
],
|
| 65 |
+
title="Image Classification with TensorFlow",
|
| 66 |
+
description="Upload an image to predict its category, price, and subcategories."
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
# Lancer l'interface
|
| 70 |
+
interface.launch()
|