Switch SAM3 Objectives to the model bundle manifest

src/dual_arm_sim/objectives/segment_image_from_no_negative_text_prompt_subtree.xml

@@ -19,10 +19,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="{confidence_threshold}"
         target_image="{image}"
         text_prompt="{text_prompt}"
@@ -45,29 +42,13 @@
         <Metadata subcategory="Perception - ML" />
       </MetadataFields>
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="{image_topic_name}" />
       <inout_port
         name="masks_visualization_topic"
         default="/masks_visualization"
       />
       <inout_port name="masks2d" default="{masks2d}" />
       <inout_port name="model_package" default="moveit_pro_sam3" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
     </SubTree>
   </TreeNodesModel>

src/dual_arm_sim/objectives/segment_point_cloud_from_text.xml

@@ -25,10 +25,6 @@
         _collapsed="true"
         masks_visualization_topic="{masks_visualization_topic}"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
         text_prompt="{text_prompt}"
         confidence_threshold="{confidence_threshold}"
         image_topic_name="{image_topic_name}"
@@ -86,18 +82,6 @@
         default="/wrist_camera/camera_info"
       />
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="/wrist_camera/color" />
       <inout_port
         name="masks_visualization_topic"
@@ -108,10 +92,6 @@
       <inout_port name="point_cloud" default="{point_cloud}" />
       <inout_port name="point_cloud_vector" default="{point_cloud_vector}" />
       <inout_port name="points_topic_name" default="/wrist_camera/points" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
       <inout_port name="use_latest_transform" default="false" />
     </SubTree>

src/hangar_sim/objectives/calibrate_sam3_mask_areas.xml

@@ -39,10 +39,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="moveit_pro_sam3"
-        encoder_model_path="models/sam3_vision_encoder.onnx"
-        decoder_model_path="models/sam3_decoder.onnx"
-        geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
-        text_encoder_model_path="models/sam3_text_encoder.onnx"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="{threshold}"
         target_image="{image}"
         text_prompt="{prompt}"

src/hangar_sim/objectives/move_boxes_to_loading_zone_from_waypoint.xml

@@ -105,10 +105,7 @@
         <Action
           ID="GetMasks2DFromExemplar"
           model_package="moveit_pro_sam3"
-          encoder_model_path="models/sam3_vision_encoder.onnx"
-          decoder_model_path="models/sam3_decoder.onnx"
-          geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
-          text_encoder_model_path="models/sam3_text_encoder.onnx"
+          model_bundle_manifest="models/model.yaml"
           confidence_threshold="{confidence_threshold}"
           target_image="{image}"
           text_prompt="{text_prompt}"

src/kitchen_sim/objectives/segment_image_from_prompt.xml

@@ -17,10 +17,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="moveit_pro_sam3"
-        encoder_model_path="models/sam3_vision_encoder.onnx"
-        decoder_model_path="models/sam3_decoder.onnx"
-        geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
-        text_encoder_model_path="models/sam3_text_encoder.onnx"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="0.5"
         target_image="{image}"
         text_prompt="a stove circular burner"

src/lab_sim/objectives/ml_find_bottles_on_table_from_image_exemplar.xml

@@ -31,10 +31,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="moveit_pro_sam3"
-        encoder_model_path="models/sam3_vision_encoder.onnx"
-        decoder_model_path="models/sam3_decoder.onnx"
-        geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
-        text_encoder_model_path="models/sam3_text_encoder.onnx"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="0.6"
         confidence_scores_str="{confidence_scores_str}"
         target_image="{wrist_camera_image}"

src/lab_sim/objectives/ml_segment_bottles_from_file.xml

@@ -14,10 +14,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="moveit_pro_sam3"
-        encoder_model_path="models/sam3_vision_encoder.onnx"
-        decoder_model_path="models/sam3_decoder.onnx"
-        geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
-        text_encoder_model_path="models/sam3_text_encoder.onnx"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="0.5"
         confidence_scores_str="{confidence_scores_str}"
         target_image="{image}"

src/lab_sim/objectives/segment_bottle_subtree.xml

@@ -60,14 +60,11 @@
           confidence_scores="{confidence_scores}"
           target_image="{image}"
           text_prompt="{object_prompt}"
-          geometry_encoder_model_path="models/sam3_geometry_encoder.onnx"
           confidence_scores_str="{confidence_scores_str}"
           mask_count="{mask_count}"
           masks2d="{refined_masks}"
           confidence_threshold="0.5"
-          text_encoder_model_path="models/sam3_text_encoder.onnx"
-          encoder_model_path="models/sam3_vision_encoder.onnx"
-          decoder_model_path="models/sam3_decoder.onnx"
+          model_bundle_manifest="models/model.yaml"
           model_package="moveit_pro_sam3"
           exemplar_image="{exemplar_image}"
           target_bboxes="{target_bboxes}"

src/lab_sim/test/objectives_integration_test.py

@@ -279,23 +279,12 @@ def _controller_safe_mujoco_reset(
     "ML Find Bottles on Table from Image Exemplar",  # Skipped because it looks for a file on a home path
     "ML Segment Image",
     "ML Segment Image Loop",
-    # Temporarily skipped: reaches GetMasks2DFromExemplar, whose four per-model path ports
-    # (encoder_model_path, text_encoder_model_path, geometry_encoder_model_path,
-    # decoder_model_path) were replaced by model_bundle_manifest + runtime_id in
-    # moveit_pro#21155. The objective still sets the removed ports, so it cannot load against
-    # that image. Re-enabled by PR #802, which performs the port swap.
-    "ML Segment Bottles from File",
     "ML Segment Point Cloud",  # Requires GPU for ONNX inference; falls back to CPU and times out waiting for /wrist_camera/points on CI runners without a camera warmup delay.
     "ML Segment Point Cloud from Clicked Point",
     "MPC Pose Tracking",
     "MPC Pose Tracking With Point Cloud Avoidance",
     "Octomap Example",  # Requires user input to clear the octomap.
     "Pick 1 Pill Bottle with ML",
-    # Temporarily skipped for the same reason as "ML Segment Bottles from File": it reaches
-    # GetMasks2DFromExemplar indirectly, through "Get Bottle Grasp via ICP Subtree" ->
-    # "Segment Bottle Subtree", and that subtree sets the removed per-model path ports.
-    # Re-enabled by PR #802.
-    "Pick 1 Pill Bottle with SAM3",
     "Pick All Bottles with AprilTags",
     "Pick All Pill Bottles",
     "Pick up Object",

src/moveit_pro_franka_configs/franka_base_config/objectives/segment_image_from_no_negative_text_prompt_subtree.xml

@@ -19,10 +19,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="{confidence_threshold}"
         target_image="{image}"
         text_prompt="{text_prompt}"
@@ -45,29 +42,13 @@
         <Metadata subcategory="Perception - ML" />
       </MetadataFields>
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="{image_topic_name}" />
       <inout_port
         name="masks_visualization_topic"
         default="/masks_visualization"
       />
       <inout_port name="masks2d" default="{masks2d}" />
       <inout_port name="model_package" default="moveit_pro_sam3" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
     </SubTree>
   </TreeNodesModel>

src/moveit_pro_ur_configs/picknik_ur_base_config/objectives/segment_image_from_no_negative_text_prompt_subtree.xml

@@ -19,10 +19,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="{confidence_threshold}"
         target_image="{image}"
         text_prompt="{text_prompt}"
@@ -45,29 +42,13 @@
         <Metadata subcategory="Perception - ML" />
       </MetadataFields>
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="{image_topic_name}" />
       <inout_port
         name="masks_visualization_topic"
         default="/masks_visualization"
       />
       <inout_port name="masks2d" default="{masks2d}" />
       <inout_port name="model_package" default="moveit_pro_sam3" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
     </SubTree>
   </TreeNodesModel>

src/moveit_pro_ur_configs/picknik_ur_base_config/objectives/segment_image_from_text_prompt_subtree.xml

@@ -19,10 +19,7 @@
       <Action
         ID="GetMasks2DFromExemplar"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
+        model_bundle_manifest="models/model.yaml"
         confidence_threshold="{confidence_threshold}"
         target_image="{image}"
         text_prompt="{text_prompt}"
@@ -43,29 +40,13 @@
         <Metadata subcategory="Perception - ML" />
       </MetadataFields>
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="{image_topic_name}" />
       <inout_port
         name="masks_visualization_topic"
         default="/masks_visualization"
       />
       <inout_port name="masks2d" default="{masks2d}" />
       <inout_port name="model_package" default="moveit_pro_sam3" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
     </SubTree>
   </TreeNodesModel>

src/moveit_pro_ur_configs/picknik_ur_base_config/objectives/segment_point_cloud_from_text_prompt_subtree.xml

@@ -25,10 +25,6 @@
         _collapsed="true"
         masks_visualization_topic="{masks_visualization_topic}"
         model_package="{model_package}"
-        encoder_model_path="{encoder_model_path}"
-        decoder_model_path="{decoder_model_path}"
-        geometry_encoder_model_path="{geometry_encoder_model_path}"
-        text_encoder_model_path="{text_encoder_model_path}"
         text_prompt="{text_prompt}"
         confidence_threshold="{confidence_threshold}"
         image_topic_name="{image_topic_name}"
@@ -85,18 +81,6 @@
         default="/wrist_camera/camera_info"
       />
       <inout_port name="confidence_threshold" default="0.5" />
-      <inout_port
-        name="decoder_model_path"
-        default="models/sam3_decoder.onnx"
-      />
-      <inout_port
-        name="encoder_model_path"
-        default="models/sam3_vision_encoder.onnx"
-      />
-      <inout_port
-        name="geometry_encoder_model_path"
-        default="models/sam3_geometry_encoder.onnx"
-      />
       <inout_port name="image_topic_name" default="/wrist_camera/color" />
       <inout_port
         name="masks_visualization_topic"
@@ -107,10 +91,6 @@
       <inout_port name="point_cloud" default="{point_cloud}" />
       <inout_port name="point_cloud_vector" default="{point_cloud_vector}" />
       <inout_port name="points_topic_name" default="/wrist_camera/points" />
-      <inout_port
-        name="text_encoder_model_path"
-        default="models/sam3_text_encoder.onnx"
-      />
       <inout_port name="text_prompt" default="{text_prompt}" />
     </SubTree>
   </TreeNodesModel>