v1

app.py

@@ -152,43 +152,49 @@ class VideoHighlightDetector:
         outputs = self.model.generate(**inputs, max_new_tokens=64, do_sample=False)
         response = self.processor.decode(outputs[0], skip_special_tokens=True).lower().split("assistant: ")[1]
         return "yes" in response
-
 def create_xspf_playlist(video_path: str, segments: list, descriptions: list) -> str:
     """Create XSPF playlist from segments with descriptions."""
-    root = ET.Element("playlist", version="1", xmlns="http://xspf.org/ns/0/")
+    # Register the VLC namespace
+    ET.register_namespace('vlc', 'http://www.videolan.org/vlc/playlist/ns/0/')
+    ET.register_namespace('', 'http://xspf.org/ns/0/')
+    
+    root = ET.Element("{http://xspf.org/ns/0/}playlist", {"version": "1"})
     
     # Get video filename for the title
     video_filename = os.path.basename(video_path)
-    title = ET.SubElement(root, "title")
+    title = ET.SubElement(root, "{http://xspf.org/ns/0/}title")
     title.text = f"{video_filename} - Highlights"
     
-    tracklist = ET.SubElement(root, "trackList")
+    tracklist = ET.SubElement(root, "{http://xspf.org/ns/0/}trackList")
     
     for idx, ((start_time, end_time), description) in enumerate(zip(segments, descriptions)):
-        track = ET.SubElement(tracklist, "track")
+        track = ET.SubElement(tracklist, "{http://xspf.org/ns/0/}track")
         
-        location = ET.SubElement(track, "location")
+        location = ET.SubElement(track, "{http://xspf.org/ns/0/}location")
         location.text = f"file:///{video_filename}"
         
-        title = ET.SubElement(track, "title")
+        title = ET.SubElement(track, "{http://xspf.org/ns/0/}title")
         title.text = f"Highlight {idx + 1}"
         
-        annotation = ET.SubElement(track, "annotation")
+        annotation = ET.SubElement(track, "{http://xspf.org/ns/0/}annotation")
         annotation.text = description
         
-        start_meta = ET.SubElement(track, "meta", rel="start")
+        start_meta = ET.SubElement(track, "{http://xspf.org/ns/0/}meta", {"rel": "start"})
         start_meta.text = format_duration(start_time)
         
-        end_meta = ET.SubElement(track, "meta", rel="end")
+        end_meta = ET.SubElement(track, "{http://xspf.org/ns/0/}meta", {"rel": "end"})
         end_meta.text = format_duration(end_time)
     
     # Add VLC extension
-    extension = ET.SubElement(root, "extension", application="http://www.videolan.org/vlc/playlist/0")
+    extension = ET.SubElement(root, "{http://xspf.org/ns/0/}extension", 
+                            {"application": "http://www.videolan.org/vlc/playlist/0"})
+    
     for i in range(len(segments)):
-        item = ET.SubElement(extension, "vlc:item", tid=str(i))
+        ET.SubElement(extension, "{http://www.videolan.org/vlc/playlist/ns/0/}item", 
+                     {"tid": str(i)})
     
     # Convert to string with pretty printing
-    xml_str = minidom.parseString(ET.tostring(root)).toprettyxml(indent="    ")
+    xml_str = minidom.parseString(ET.tostring(root, encoding='unicode')).toprettyxml(indent="    ")
     return xml_str
 
 def create_ui(examples_path: str, model_path: str):
@@ -280,6 +286,7 @@ def create_ui(examples_path: str, model_path: str):
                         
                         if detector.process_segment(temp_segment.name, highlights):
                             # Get segment description
+                            print("KEEPING SEGMENT")
                             description = detector.analyze_segment(temp_segment.name)
                             kept_segments.append((start_time, end_time))
                             segment_descriptions.append(description)