DeepLabCut · AlexEMG · Sep 10, 2025 · Jul 8, 2025 · Jul 8, 2025 · Jul 8, 2025
diff --git a/deeplabcut/gui/tabs/modelzoo.py b/deeplabcut/gui/tabs/modelzoo.py
@@ -241,7 +241,9 @@ def _build_common_attributes(self) -> None:
 
         self.model_combo.currentTextChanged.connect(self._update_pose_models)
         self.model_combo.currentTextChanged.connect(self._update_detectors)
-        self.model_combo.currentTextChanged.connect(self._update_adaptation_visibility)
+        self.model_combo.currentTextChanged.connect(
+            self._update_adaptation_detector_visibility
+        )
 
     def _add_tf_scales_row(self, layout: QtWidgets.QGridLayout):
         scales_label = QtWidgets.QLabel("Scale list")
@@ -359,8 +361,10 @@ def _add_torch_adaptation_settings_row(self, layout: QtWidgets.QGridLayout):
         self.torch_adapt_epoch_spinbox.setRange(1, 50)
         self.torch_adapt_epoch_spinbox.setValue(4)
         self.torch_adapt_epoch_spinbox.setMaximumWidth(100)
-        adapt_det_epoch_label = QtWidgets.QLabel("Number of detector adaptation epochs")
-        adapt_det_epoch_label.setMinimumWidth(200)
+        self.adapt_det_epoch_label = QtWidgets.QLabel(
+            "Number of detector adaptation epochs"
+        )
+        self.adapt_det_epoch_label.setMinimumWidth(200)
         self.torch_adapt_det_epoch_spinbox = QtWidgets.QSpinBox()
         self.torch_adapt_det_epoch_spinbox.setRange(1, 50)
         self.torch_adapt_det_epoch_spinbox.setValue(4)
@@ -374,7 +378,7 @@ def _add_torch_adaptation_settings_row(self, layout: QtWidgets.QGridLayout):
         self.torch_adaptation_settings_row.addWidget(adapt_epoch_label)
         self.torch_adaptation_settings_row.addWidget(self.torch_adapt_epoch_spinbox)
         self.torch_adaptation_settings_row.addSpacing(20)
-        self.torch_adaptation_settings_row.addWidget(adapt_det_epoch_label)
+        self.torch_adaptation_settings_row.addWidget(self.adapt_det_epoch_label)
         self.torch_adaptation_settings_row.addWidget(self.torch_adapt_det_epoch_spinbox)
         self.torch_adaptation_settings_row.addStretch()
         layout.addLayout(self.torch_adaptation_settings_row, 2, 0, 1, 6)
@@ -401,6 +405,10 @@ def _adapt_checkbox_status_changed(self, state: int) -> None:
             set_layout_contents_visible(
                 self.torch_adaptation_settings_row, Qt.CheckState(state) == Qt.Checked
             )
+            if Qt.CheckState(state) == Qt.Checked:
+                self._update_adaptation_detector_visibility(
+                    self.model_combo.currentText()
+                )
 
     def select_folder(self):
         dirname = QtWidgets.QFileDialog.getExistingDirectory(
@@ -563,11 +571,7 @@ def _gather_kwargs(self) -> dict:
             kwargs["adapt_iterations"] = self.adapt_iter_spinbox.value()
         else:
             kwargs["detector_name"] = self.detector_type_selector.currentText()
-            kwargs["video_adapt"] = (
-                self.adapt_checkbox.isChecked()
-                if self.model_combo.currentText() != "superanimal_humanbody"
-                else False
-            )
+            kwargs["video_adapt"] = (self.adapt_checkbox.isChecked())
             kwargs["pseudo_threshold"] = self.pose_threshold_spinbox.value()
             kwargs["bbox_threshold"] = self.detector_threshold_spinbox.value()
             kwargs["detector_epochs"] = self.torch_adapt_det_epoch_spinbox.value()
@@ -624,14 +628,11 @@ def _update_detectors(self, super_animal: str) -> None:
             self.detector_row, self.root.engine == Engine.PYTORCH
         )
 
-    def _update_adaptation_visibility(self, super_animal: str):
-        if (
-            self.root.engine == Engine.PYTORCH
-            and super_animal != "superanimal_humanbody"
-        ):
-            self.torch_widget.show()
-        else:
-            self.torch_widget.hide()
+    def _update_adaptation_detector_visibility(self, superanimal: str):
+        self.adapt_det_epoch_label.setVisible((superanimal != "superanimal_humanbody"))
+        self.torch_adapt_det_epoch_spinbox.setVisible(
+            (superanimal != "superanimal_humanbody")
+        )
 
     @Slot(Engine)
     def _on_engine_change(self, engine: Engine) -> None:

diff --git a/deeplabcut/modelzoo/video_inference.py b/deeplabcut/modelzoo/video_inference.py
@@ -402,12 +402,6 @@ def video_inference_superanimal(
 
         output_suffix = "_before_adapt"
 
-        if superanimal_name == "superanimal_humanbody" and video_adapt:
-            print(
-                f"Video adaptation currently not supported for {superanimal_name}. Setting it to false."
-            )
-            video_adapt = False
-
         if video_adapt:
             # the users can pass in many videos. For now, we only use one video for
             # video adaptation. As reported in Ye et al. 2024, one video should be
@@ -485,60 +479,72 @@ def video_inference_superanimal(
                 )
 
             model_snapshot_prefix = f"snapshot-{model_name}"
-            detector_snapshot_prefix = f"snapshot-{detector_name}"
-
             config["runner"]["snapshot_prefix"] = model_snapshot_prefix
-            config["detector"]["runner"]["snapshot_prefix"] = detector_snapshot_prefix
+
+            if superanimal_name != "superanimal_humanbody":
+                detector_snapshot_prefix = f"snapshot-{detector_name}"
+                config["detector"]["runner"][
+                    "snapshot_prefix"
+                ] = detector_snapshot_prefix
 
             # the model config's parameters need to be updated for adaptation training
             model_config_path = model_folder / "pytorch_config.yaml"
             with open(model_config_path, "w") as f:
                 yaml = YAML()
                 yaml.dump(config, f)
 
-            # get the current epoch of the detector and pose model
+            # get the current epoch of the pose model
             current_pose_epoch = get_checkpoint_epoch(pose_model_path)
-            current_detector_epoch = get_checkpoint_epoch(detector_path)
             # update the checkpoint path with the current epoch, if the checkpoint does not exist, use the best checkpoint
-            adapted_detector_checkpoint = (
-                model_folder
-                / f"{detector_snapshot_prefix}-{current_detector_epoch + detector_epochs:03}.pt"
-            )
             adapted_pose_checkpoint = (
                 model_folder
                 / f"{model_snapshot_prefix}-{current_pose_epoch + pose_epochs:03}.pt"
             )
-            if not Path(adapted_detector_checkpoint).exists():
-                adapted_detector_checkpoint = (
-                    model_folder
-                    / f"{detector_snapshot_prefix}-best-{current_detector_epoch + detector_epochs:03}.pt"
-                )
             if not Path(adapted_pose_checkpoint).exists():
                 adapted_pose_checkpoint = (
                     model_folder
                     / f"{model_snapshot_prefix}-best-{current_pose_epoch + pose_epochs:03}.pt"
                 )
 
+            if superanimal_name != "superanimal_humanbody":
+                current_detector_epoch = get_checkpoint_epoch(detector_path)
+                adapted_detector_checkpoint = (
+                    model_folder
+                    / f"{detector_snapshot_prefix}-{current_detector_epoch + detector_epochs:03}.pt"
+                )
+                if not Path(adapted_detector_checkpoint).exists():
+                    adapted_detector_checkpoint = (
+                        model_folder
+                        / f"{detector_snapshot_prefix}-best-{current_detector_epoch + detector_epochs:03}.pt"
+                    )
+
             if (
-                adapted_detector_checkpoint.exists()
-                and adapted_pose_checkpoint.exists()
-            ):
+                superanimal_name == "superanimal_humanbody"
+                or adapted_detector_checkpoint.exists()
+            ) and adapted_pose_checkpoint.exists():
+                snapshots_msg = f"pose ({adapted_pose_checkpoint})"
+                if superanimal_name != "superanimal_humanbody":
+                    snapshots_msg += f" and detector ({adapted_detector_checkpoint})"
                 print(
-                    f"Video adaptation already ran; pose ({adapted_pose_checkpoint}) "
-                    f"and detector ({adapted_detector_checkpoint}) already exist. To "
-                    "rerun video adaptation training, delete the checkpoints or select"
-                    "a different number of adaptation epochs. Continuing with the"
-                    "existing checkpoints."
+                    f"Video adaptation already ran; {snapshots_msg} already exist. "
+                    "To rerun video adaptation training, delete the checkpoints or select a different "
+                    "number of adaptation epochs. Continuing with the existing checkpoints."
                 )
             else:
-                print(
-                    "Running video adaptation with following parameters:\n"
+                params_msg = (
+                    f"  video adaptation batch size: {video_adapt_batch_size}\n"
                     f"  (pose training) pose_epochs: {pose_epochs}\n"
                     "  (pose) save_epochs: 1\n"
-                    f"  detector_epochs: {detector_epochs}\n"
-                    "  detector_save_epochs: 1\n"
-                    f"  video adaptation batch size: {video_adapt_batch_size}\n"
                 )
+                if superanimal_name != "superanimal_humanbody":
+                    params_msg += (
+                        f"  detector_epochs: {detector_epochs}\n"
+                        "  detector_save_epochs: 1\n"
+                    )
+                print(
+                    "Running video adaptation with following parameters:\n" + params_msg
+                )
+
                 train_file = pseudo_dataset_folder / "annotations" / "train.json"
                 with open(train_file, "r") as f:
                     temp_obj = json.load(f)
@@ -551,6 +557,11 @@ def video_inference_superanimal(
                     )
                     return
 
+                if superanimal_name == "superanimal_humanbody":
+                    print(
+                        "Warning, with the superanimal_humanbody type, only the pose model is adapted"
+                    )
+
                 adaptation_train(
                     project_root=pseudo_dataset_folder,
                     model_folder=model_folder,
@@ -566,32 +577,35 @@ def video_inference_superanimal(
                     detector_path=detector_path,
                     batch_size=video_adapt_batch_size,
                     detector_batch_size=video_adapt_batch_size,
+                    skip_detector=(superanimal_name == "superanimal_humanbody"),
                 )
 
             # after video adaptation, re-update the adapted checkpoint path, if the checkpoint does not exist, use the best checkpoint
-            adapted_detector_checkpoint = (
-                model_folder
-                / f"{detector_snapshot_prefix}-{current_detector_epoch + detector_epochs:03}.pt"
-            )
             adapted_pose_checkpoint = (
                 model_folder
                 / f"{model_snapshot_prefix}-{current_pose_epoch + pose_epochs:03}.pt"
             )
-            if not Path(adapted_detector_checkpoint).exists():
-                adapted_detector_checkpoint = (
-                    model_folder
-                    / f"{detector_snapshot_prefix}-best-{current_detector_epoch + detector_epochs:03}.pt"
-                )
             if not Path(adapted_pose_checkpoint).exists():
                 adapted_pose_checkpoint = (
                     model_folder
                     / f"{model_snapshot_prefix}-best-{current_pose_epoch + pose_epochs:03}.pt"
                 )
+            pose_model_path = adapted_pose_checkpoint
+
+            if superanimal_name != "superanimal_humanbody":
+                adapted_detector_checkpoint = (
+                    model_folder
+                    / f"{detector_snapshot_prefix}-{current_detector_epoch + detector_epochs:03}.pt"
+                )
+                if not Path(adapted_detector_checkpoint).exists():
+                    adapted_detector_checkpoint = (
+                        model_folder
+                        / f"{detector_snapshot_prefix}-best-{current_detector_epoch + detector_epochs:03}.pt"
+                    )
+                detector_path = adapted_detector_checkpoint
 
             # Set the customized checkpoint paths and
             output_suffix = "_after_adapt"
-            detector_path = adapted_detector_checkpoint
-            pose_model_path = adapted_pose_checkpoint
 
         return _video_inference_superanimal(
             videos,

diff --git a/deeplabcut/modelzoo/weight_initialization.py b/deeplabcut/modelzoo/weight_initialization.py
@@ -15,7 +15,7 @@
 from deeplabcut.core.config import read_config_as_dict
 from deeplabcut.core.weight_init import WeightInitialization
 from deeplabcut.pose_estimation_pytorch.modelzoo.utils import (
-    get_super_animal_snapshot_path
+    get_super_animal_snapshot_path,
 )
 
 
@@ -71,6 +71,11 @@ def build_weight_init(
     Returns:
         The built WeightInitialization.
     """
+    if super_animal == "superanimal_humanbody":
+        raise NotImplementedError(
+            "Weight Initialization, Transfer-Learning and Finetuning is currently not supported for superanimal_humanbody"
+        )
+
     if isinstance(cfg, (str, Path)):
         cfg = read_config_as_dict(cfg)
 

diff --git a/deeplabcut/pose_estimation_pytorch/modelzoo/train_from_coco.py b/deeplabcut/pose_estimation_pytorch/modelzoo/train_from_coco.py
@@ -37,6 +37,7 @@ def adaptation_train(
     batch_size: int = 8,
     detector_batch_size: int = 8,
     eval_interval: int | None = None,
+    skip_detector: bool = False,
 ):
     setup_file_logging(Path(model_folder) / "log.txt")
     loader = COCOLoader(
@@ -49,25 +50,30 @@ def adaptation_train(
     utils.fix_seeds(loader.model_cfg["train_settings"]["seed"])
 
     updates = {
-        "detector.model.freeze_bn_stats": True,
-        "detector.runner.snapshots.max_snapshots": 5,
-        "detector.runner.snapshots.save_epochs": detector_save_epochs or 1,
-        "detector.train_settings.batch_size": detector_batch_size,
-        "detector.train_settings.epochs": detector_epochs or 4,
         "model.backbone.freeze_bn_stats": True,
         "runner.snapshots.max_snapshots": 5,
         "runner.snapshots.save_epochs": save_epochs or 1,
         "train_settings.batch_size": batch_size,
         "train_settings.epochs": epochs or 4,
     }
+    if not skip_detector:
+        updates.update(
+            {
+                "detector.model.freeze_bn_stats": True,
+                "detector.runner.snapshots.max_snapshots": 5,
+                "detector.runner.snapshots.save_epochs": detector_save_epochs or 1,
+                "detector.train_settings.batch_size": detector_batch_size,
+                "detector.train_settings.epochs": detector_epochs or 4,
+            }
+        )
 
     if eval_interval is not None:
         updates["runner.eval_interval"] = eval_interval
 
     loader.update_model_cfg(updates)
 
     pose_task = Task(loader.model_cfg["method"])
-    if pose_task == Task.TOP_DOWN:
+    if pose_task == Task.TOP_DOWN and not skip_detector:
         logger_config = None
         if loader.model_cfg.get("logger"):
             logger_config = copy.deepcopy(loader.model_cfg["logger"])