Upload 2 files

preprocessing_router.py

@@ -492,9 +492,25 @@ def process_video_job(job_id: str):
                             fn = f"face_{frame_idx:06d}_{saved_count:03d}.jpg"
                             local_crop_path = faces_root / fn
                             
-                            if crop_path and os.path.exists(crop_path):
-                                shutil.copy2(crop_path, local_crop_path)
-                            else:
+                            crop_saved = False
+                            if crop_path:
+                                # Handle remote URLs from svision (Gradio)
+                                if isinstance(crop_path, str) and crop_path.startswith("http"):
+                                    try:
+                                        import requests
+                                        resp = requests.get(crop_path, timeout=30)
+                                        if resp.status_code == 200:
+                                            with open(local_crop_path, "wb") as f:
+                                                f.write(resp.content)
+                                            crop_saved = True
+                                    except Exception as dl_err:
+                                        print(f"[{job_id}] Error descargando crop: {dl_err}")
+                                # Handle local paths
+                                elif isinstance(crop_path, str) and os.path.exists(crop_path):
+                                    shutil.copy2(crop_path, local_crop_path)
+                                    crop_saved = True
+                            
+                            if not crop_saved:
                                 # If no crop from svision, use original frame
                                 shutil.copy2(frame_path, local_crop_path)
                             
@@ -536,8 +552,18 @@ def process_video_job(job_id: str):
             chars_dir = base / "characters"
             chars_dir.mkdir(parents=True, exist_ok=True)
 
+            print(f"[{job_id}] cluster_map: {cluster_map}")
+            print(f"[{job_id}] crops_meta count: {len(crops_meta)}")
+            print(f"[{job_id}] faces_root: {faces_root}, exists: {faces_root.exists()}")
+            if faces_root.exists():
+                existing_files = list(faces_root.glob("*"))
+                print(f"[{job_id}] Files in faces_root: {len(existing_files)}")
+                for ef in existing_files[:5]:
+                    print(f"[{job_id}]   - {ef.name}")
+            
             for ci, idxs in sorted(cluster_map.items(), key=lambda x: x[0]):
                 char_id = f"char_{ci:02d}"
+                print(f"[{job_id}] Processing cluster {char_id} with {len(idxs)} indices: {idxs[:5]}...")
                 
                 if not idxs:
                     continue
@@ -555,10 +581,12 @@ def process_video_job(job_id: str):
                 
                 for j in selected_idxs:
                     if j >= len(crops_meta):
+                        print(f"[{job_id}]   Index {j} out of range (crops_meta len={len(crops_meta)})")
                         continue
                     meta = crops_meta[j]
                     fname = meta.get("file")
                     if not fname:
+                        print(f"[{job_id}]   No filename in meta for index {j}")
                         continue
                     
                     src = faces_root / fname
@@ -568,8 +596,10 @@ def process_video_job(job_id: str):
                             shutil.copy2(src, dst)
                             files.append(fname)
                             file_urls.append(f"/files/{video_name}/{char_id}/{fname}")
-                    except Exception:
-                        pass
+                        else:
+                            print(f"[{job_id}]   Source file not found: {src}")
+                    except Exception as cp_err:
+                        print(f"[{job_id}]   Error copying {fname}: {cp_err}")
 
                 # Create representative image
                 rep = files[0] if files else None

svision_client.py

@@ -3,6 +3,7 @@ os.environ["CUDA_VISIBLE_DEVICES"] = "1"
 
 from gradio_client import Client, handle_file
 from typing import Any, Dict, List, Optional, Tuple, Union
+import requests
 import json
 
 # Lazy initialization to avoid crash if Space is down at import time
@@ -123,6 +124,20 @@ def extract_descripcion_escena(imagen_path: str) -> str:
     return result
 
 
+def _extract_path_from_gradio_file(file_obj) -> Optional[str]:
+    """Extract file path from Gradio file object (can be dict, str, or other)."""
+    if file_obj is None:
+        return None
+    if isinstance(file_obj, str):
+        return file_obj
+    if isinstance(file_obj, dict):
+        # Gradio returns dicts like {"path": "...", "url": "...", "orig_name": "..."}
+        return file_obj.get("path") or file_obj.get("url") or file_obj.get("name")
+    if hasattr(file_obj, "name"):
+        return file_obj.name
+    return str(file_obj)
+
+
 def get_face_embeddings_from_image(image_path: str) -> List[Dict[str, Any]]:
     """
     Call the /face_image_embedding_casting endpoint to detect faces and get embeddings.
@@ -137,31 +152,44 @@ def get_face_embeddings_from_image(image_path: str) -> List[Dict[str, Any]]:
     Returns
     -------
     List[Dict[str, Any]]
-        List of dicts with 'embedding' (list of floats) and 'face_crop' (image path).
+        List of dicts with 'embedding' (list of floats) and 'face_crop_path' (image path string).
         Returns empty list if no faces detected or on error.
     """
     try:
-        # Returns: (face_crops: list of images, face_embeddings: list of dicts)
+        # Returns: (face_crops: list of images/dicts, face_embeddings: list of dicts)
         result = _get_svision_client().predict(
             image=handle_file(image_path),
             api_name="/face_image_embedding_casting"
         )
-        # result is a tuple: (list of image paths, list of embedding dicts)
+        
+        # result is a tuple: (list of image paths/dicts, list of embedding dicts)
         if result and len(result) >= 2:
-            face_crops = result[0] if result[0] else []
+            face_crops_raw = result[0] if result[0] else []
             face_embeddings = result[1] if result[1] else []
-            # Combine into unified structure
+            
+            # Combine into unified structure, extracting paths correctly
             faces = []
             for i, emb_dict in enumerate(face_embeddings):
+                # Extract path from Gradio file object (might be dict or string)
+                crop_path = None
+                if i < len(face_crops_raw):
+                    crop_path = _extract_path_from_gradio_file(face_crops_raw[i])
+                
+                embedding = emb_dict.get("embedding", []) if isinstance(emb_dict, dict) else []
+                
                 faces.append({
-                    "embedding": emb_dict.get("embedding", []),
-                    "face_crop_path": face_crops[i] if i < len(face_crops) else None,
-                    "index": emb_dict.get("index", i),
+                    "embedding": embedding,
+                    "face_crop_path": crop_path,
+                    "index": emb_dict.get("index", i) if isinstance(emb_dict, dict) else i,
                 })
+            
+            print(f"[svision_client] Detected {len(faces)} faces from image")
             return faces
         return []
     except Exception as e:
         print(f"[svision_client] get_face_embeddings_from_image error: {e}")
+        import traceback
+        traceback.print_exc()
         return []