Remove NaN fallback for empty video predictions

Simplifies `analyze_videos` by always passing the inference `predictions` directly to `create_df_from_prediction` when `save_as_df` is enabled, and removes the special-case branch that generated all-NaN placeholder rows when no predictions were produced. The unused `_nan_predictions` helper and its warning/error handling were also deleted.
C-Achard · Sep 8, 2026 · Sep 8, 2026 · Sep 8, 2026 · Sep 9, 2026 · Sep 9, 2026
commit bff4fea56b11101e9bf4e58ab8af140c087ef022
diff --git a/deeplabcut/pose_estimation_pytorch/apis/videos.py b/deeplabcut/pose_estimation_pytorch/apis/videos.py
@@ -614,37 +614,21 @@ def analyze_videos(
                 print("Can't ``save_as_df`` as ``use_shelve=True``. Skipping.")
 
             if not use_shelve:
-                output_data = _generate_output_data(pose_cfg, predictions)  # this should not consume full-NaN frames
+                output_data = _generate_output_data(pose_cfg, predictions)
                 with Path(output_pkl).open("wb") as f:
                     pickle.dump(output_data, f, pickle.HIGHEST_PROTOCOL)
 
                 if save_as_df:
-                    df_predictions = predictions
-                    if not df_predictions:
-                        n_frames = video_iterator.get_n_frames(robust=robust_nframes)
-                        logging.warning(
-                            f"No predictions were produced for {video}: writing an "
-                            f"all-NaN results file with {n_frames} row(s) so that "
-                            "downstream steps proceed."
-                        )
-                        df_predictions = _nan_predictions(n_frames, loader.model_cfg)
-
-                    if not df_predictions:
-                        logging.error(
-                            f"Cannot write a results file for {video}: no predictions "
-                            "were produced and the video reports 0 frames."
-                        )
-                    else:
-                        create_df_from_prediction(
-                            predictions=df_predictions,
-                            multi_animal=multi_animal,
-                            model_cfg=loader.model_cfg,
-                            dlc_scorer=dlc_scorer,
-                            output_path=output_path,
-                            output_prefix=output_prefix,
-                            save_as_csv=save_as_csv,
-                        )
-                        h5_files_created = True
+                    create_df_from_prediction(
+                        predictions=predictions,
+                        multi_animal=multi_animal,
+                        model_cfg=loader.model_cfg,
+                        dlc_scorer=dlc_scorer,
+                        output_path=output_path,
+                        output_prefix=output_prefix,
+                        save_as_csv=save_as_csv,
+                    )
+                    h5_files_created = True  # .h5 file was created
 
             if multi_animal:
                 assemblies_path = output_path / f"{output_prefix}_assemblies.pickle"
@@ -804,28 +788,6 @@ def create_df_from_prediction(
     return df
 
 
-def _nan_predictions(n_frames: int, model_cfg: dict) -> list[dict[str, np.ndarray]]:
-    """Returns placeholder all-NaN predictions, one per frame.
-
-    Used when inference produced no predictions at all, so a well-formed results
-    file with one row per frame is still written and downstream steps (filtering,
-    labeled videos) keep working.
-    """
-    metadata = model_cfg["metadata"]
-    num_idv = len(metadata["individuals"])
-    num_bpts = len(metadata["bodyparts"])
-    num_unique = len(metadata["unique_bodyparts"])
-
-    predictions = []
-    for _ in range(n_frames):
-        prediction = dict(bodyparts=np.full((num_idv, num_bpts, 3), np.nan))
-        if num_unique > 0:
-            prediction["unique_bodyparts"] = np.full((1, num_unique, 3), np.nan)
-        predictions.append(prediction)
-
-    return predictions
-
-
 def _generate_assemblies_file(
     full_data_path: Path,
     output_path: Path,