� Ajout de pandas dans requirements + export CSV dans app.py

app.py

@@ -1,4 +1,4 @@
-# app.py – Hugging Face Spaces avec liste dynamique depuis EC2
+# app.py – Hugging Face Spaces avec liste dynamique depuis EC2 + export CSV
 
 import requests
 from requests.exceptions import RequestException
@@ -8,6 +8,7 @@ from PIL import Image
 import io
 import numpy as np
 from collections import deque
+import pandas as pd
 
 API_URL = "http://18.212.167.3:8000/predict"
 IMAGE_BASE_URL = "http://18.212.167.3:8000/test_images/"
@@ -15,13 +16,14 @@ IMAGE_LIST_URL = "http://18.212.167.3:8000/list-test-images"
 
 legend_img = Image.new("RGB", (300, 50), color=(128, 128, 128))
 history = deque(maxlen=5)
+history_csv = []
 
 def overlay_mask_on_image(image: Image.Image, mask: Image.Image, alpha: float = 0.5) -> Image.Image:
     image = image.convert("RGBA").resize(mask.size)
     mask = mask.convert("RGBA")
     return Image.blend(image, mask, alpha=alpha)
 
-def segment_image(image: Image.Image):
+def segment_image(image: Image.Image, source_name="image_upload.png"):
     try:
         if image is None:
             raise ValueError("Aucune image fournie.")
@@ -37,6 +39,13 @@ def segment_image(image: Image.Image):
         mask_bytes = base64.b64decode(data["mask_base64"])
         mask_image = Image.open(io.BytesIO(mask_bytes)).convert("RGB")
         overlay_img = overlay_mask_on_image(image, mask_image)
+
+        # Ajout dans l'historique CSV
+        history_csv.append({
+            "filename": source_name,
+            "inference_time": round(float(data["inference_time"]), 3)
+        })
+
         return image, mask_image, overlay_img, f"{data['inference_time']} sec"
 
     except RequestException as e:
@@ -66,6 +75,13 @@ def load_image_from_url(filename):
         print(f"[ERREUR CHARGEMENT IMAGE {filename}] {e}")
         return None
 
+def export_csv():
+    if not history_csv:
+        return gr.File.update(value=None, visible=False)
+    df = pd.DataFrame(history_csv)
+    csv_bytes = df.to_csv(index=False).encode()
+    return gr.File.update(value=io.BytesIO(csv_bytes), visible=True, label="📥 Résultats CSV")
+
 with gr.Blocks(title="Segmentation d'Images Urbaines") as demo:
     gr.Markdown("# 🧠 Segmentation d'Images Urbaines")
     gr.Markdown("Upload une image ou sélectionne une image distante depuis EC2")
@@ -78,14 +94,20 @@ with gr.Blocks(title="Segmentation d'Images Urbaines") as demo:
         img_mask = gr.Image(label="Mask prédit")
         img_overlay = gr.Image(label="Superposition (image + mask)")
         inf_time = gr.Textbox(label="Temps d'inférence")
-    btn.click(fn=segment_image, inputs=input_image, outputs=[img_original, img_mask, img_overlay, inf_time])
+    btn.click(fn=lambda img: segment_image(img), inputs=input_image,
+              outputs=[img_original, img_mask, img_overlay, inf_time])
 
     gr.Markdown("## 🌍 Ou sélectionne une image hébergée sur EC2 :")
     with gr.Row():
         dropdown = gr.Dropdown(label="Fichier distant (EC2)", choices=get_remote_image_names())
         btn_url = gr.Button("Segmenter cette image distante")
-    btn_url.click(fn=lambda name: segment_image(load_image_from_url(name)), inputs=dropdown,
-                  outputs=[img_original, img_mask, img_overlay, inf_time])
+    btn_url.click(fn=lambda name: segment_image(load_image_from_url(name), source_name=name),
+                  inputs=dropdown, outputs=[img_original, img_mask, img_overlay, inf_time])
+
+    with gr.Row():
+        btn_export = gr.Button("📤 Exporter CSV des inférences")
+        file_output = gr.File(visible=False)
+        btn_export.click(fn=export_csv, inputs=[], outputs=file_output)
 
     with gr.Accordion("🎨 Légende des classes", open=False):
         gr.Image(value=legend_img, label="Légende (fictive ici)", interactive=False)

requirements.txt

@@ -4,4 +4,5 @@ matplotlib-inline==0.1.7
 numpy==1.26.4
 pandas==2.2.1
 pillow==10.3.0
-requests==2.32.3
+requests==2.32.3
+pandas==2.2.1