Update app.py

app.py

@@ -1,13 +1,15 @@
 import gradio as gr
 from transformers import pipeline
 import cv2
+from PIL import Image
+import io
 
 def video_to_descriptions(video):
     # Charger le modèle via pipeline
-    model = pipeline('image-to-text', model='nlpconnect/vit-gpt2-image-captioning')
+    model = pipeline('text2text-generation', model='nlpconnect/vit-gpt2-image-captioning')
     
     # Ouvrir la vidéo
-    cap = cv2.VideoCapture(video)
+    cap = cv2.VideoCapture(video.name)
     fps = int(cap.get(cv2.CAP_PROP_FPS))
     
     descriptions = []
@@ -22,8 +24,10 @@ def video_to_descriptions(video):
         if frame_count % (fps // 2) == 0:
             # Convertir l'image en RGB
             frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            # Convertir le tableau numpy en une image PIL
+            pil_img = Image.fromarray(frame_rgb)
             # Obtenir la description de l'image
-            outputs = model(frame_rgb)
+            outputs = model(pil_img)
             description = outputs[0]['describe-text']
             descriptions.append(description)