Show catalog gallery immediately after video upload

frame_extraction/src/frame_extraction/app.py

@@ -51,11 +51,12 @@ def summarize_catalog(catalog_path: Path) -> Tuple[str, list[dict[str, Any]], li
     return message, index, gallery
 
 
-def build_catalog_from_video(file: gr.FileData) -> tuple[str | None, str, list[dict[str, Any]], list[list[str]]]:
-    if file is None:
+def build_catalog_from_video(files: list[gr.FileData]) -> tuple[str | None, str, list[dict[str, Any]], list[list[str]], list[dict[str, Any]], list[list[str]]]:
+    if not files:
         raise gr.Error("Please upload a source video first.")
 
     ensure_output_dirs()
+    file = files[0]
     run_id = uuid.uuid4().hex[:8]
     video_dir = OUTPUT_DIR / "videos"
     video_path = video_dir / f"{run_id}_{Path(file.name).name}"
@@ -65,14 +66,17 @@ def build_catalog_from_video(file: gr.FileData) -> tuple[str | None, str, list[d
     cfg = CatalogConfig(video_path=video_path, output_dir=catalog_dir)
     catalog_path = build_catalog(cfg)
     message, index, gallery = summarize_catalog(catalog_path)
-    return str(catalog_path), message, index, gallery
+    # When a new catalog is generated, clear existing matches display
+    empty_matches: list[dict[str, Any]] = []
+    empty_gallery: list[list[str]] = []
+    return str(catalog_path), message, index, gallery, empty_matches, empty_gallery
 
 
-def predict_from_arrays(arrays: list[np.ndarray], catalog_path: str | None) -> tuple[list[dict[str, Any]], list[list[str]]]:
+def predict_from_arrays(files: list[gr.FileData], catalog_path: str | None) -> tuple[list[dict[str, Any]], list[list[str]]]:
     if not catalog_path:
         raise gr.Error("Catalog not ready yet. Upload a video first.")
 
-    if not arrays:
+    if not files:
         raise gr.Error("Please upload at least one frame.")
 
     ensure_output_dirs()
@@ -80,7 +84,10 @@ def predict_from_arrays(arrays: list[np.ndarray], catalog_path: str | None) -> t
     frames_dir = OUTPUT_DIR / "frames" / run_id
     frames_dir.mkdir(parents=True, exist_ok=True)
 
-    for idx, array in enumerate(arrays):
+    arrays: list[np.ndarray] = []
+    for idx, file in enumerate(files):
+        array = np.array(Image.open(file.name).convert("RGB"))
+        arrays.append(array)
         Image.fromarray(array).save(frames_dir / f"upload_{idx:03d}.png")
 
     output_path = OUTPUT_DIR / f"matches_{run_id}.json"
@@ -119,7 +126,11 @@ def build_interface() -> gr.Blocks:
         catalog_json = gr.JSON(label="Character Index", value=initial_index)
         catalog_gallery = gr.Gallery(label="Catalog Characters", columns=4, value=initial_gallery)
 
-        video_upload = gr.File(label="Source video", file_types=["video"], height="auto")
+        video_upload = gr.UploadButton(
+            label="Upload source video",
+            file_types=["video"],
+            file_count="single",
+        )
         frame_upload = gr.UploadButton(
             label="Upload frames",
             file_types=["image"],
@@ -129,16 +140,17 @@ def build_interface() -> gr.Blocks:
         matches_json = gr.JSON(label="Matches")
         match_gallery = gr.Gallery(label="Matched Characters", columns=3)
 
-        def on_video_upload(file: gr.FileData) -> tuple[str | None, str, list[dict[str, Any]], list[list[str]]]:
-            return build_catalog_from_video(file)
-
-        video_upload.change(on_video_upload, inputs=video_upload, outputs=[catalog_state, status_box, catalog_json, catalog_gallery])
-
-        def on_frames_upload(files: list[gr.FileData], catalog_path: str | None) -> tuple[list[dict[str, Any]], list[list[str]]]:
-            arrays = [np.array(Image.open(file.name).convert("RGB")) for file in files]
-            return predict_from_arrays(arrays, catalog_path)
+        video_upload.upload(
+            build_catalog_from_video,
+            inputs=video_upload,
+            outputs=[catalog_state, status_box, catalog_json, catalog_gallery, matches_json, match_gallery],
+        )
 
-        frame_upload.upload(on_frames_upload, inputs=[frame_upload, catalog_state], outputs=[matches_json, match_gallery])
+        frame_upload.upload(
+            predict_from_arrays,
+            inputs=[frame_upload, catalog_state],
+            outputs=[matches_json, match_gallery],
+        )
     return demo