Open Vocabulary Refactor

.gitignore

@@ -105,3 +105,9 @@ venv.bak/
 
 #idea
 .idea/
+
+# vscode
+.vscode/
+
+# log
+log/

README.md

@@ -11,3 +11,55 @@ This project depends on ROS [vision_msgs](http://wiki.ros.org/vision_msgs), whic
 You can select a specific model using the `model_file` ros parameter, which specifies a given `.yaml` file under `detectron2/configs`. Detectron2 is then initialized with that model, and you can send it images to be segmented through the `/detectron/segment` service. See the [service message specification](segmentation_msgs/srv/SegmentImage.srv) for details about the output format. 
 
 If the `publish_visualization` parameter is set to true, the node also publishes a version of image annotated with the masks and scores, on a topic specified through the `visualization_topic` parameter.
+
+## Usage with [Voxeland](https://github.com/MAPIRlab/Voxeland)
+
+### 1. Build the Workspace
+
+Clean previous build artifacts and execute `colcon build` to compile the workspace. Please adapt the paths if necessary:
+
+```bash
+cd ~/ros2_ws
+rm -rf build/ install/ log/
+colcon build --symlink-install --cmake-clean-cache
+```
+
+### 2. Launch the Detectron2 ROS 2 Node
+
+Please adapt the paths if necessary:
+
+```bash
+cd ~/ros2_ws
+source install/setup.bash
+ros2 run detectron_ros detectron_ros_node
+```
+
+### 3. Run Voxeland and Play a ScanNet ROS Bag
+
+Create and execute a bash script that contains the following commands. Please adapt the paths if necessary:
+
+```bash
+cd ~/ros2_ws
+
+# Init voxeland_robot_perception with Detectron2 detector  
+gnome-terminal -- bash -c "source ~/.bashrc; source /home/ubuntu/ros2_ws/venvs/voxenv/bin/activate; ros2 launch voxeland_robot_perception semantic_mapping.launch.py object_detector:=detectron; exec bash"
+
+# Init voxeland server
+gnome-terminal -- bash -c "ros2 launch voxeland voxeland_server.launch.xml; exec bash"
+
+# Open bag folder and play ros2 bag
+gnome-terminal -- bash -c "cd /home/ubuntu/ros2_ws/bag/ScanNet/to_ros/ROS2_bags/scene0000_01/; ros2 bag play scene0000_01.db3; exec bash"
+```
+
+### Service Interface
+
+**Service:** `/yoloe/segment`  
+**Type:** `segmentation_msgs/srv/SegmentImage`
+
+**Request:**
+- `sensor_msgs/Image image` - Input RGB image
+
+**Response:**
+- `segmentation_msgs/SemanticInstance2D[] instances` - Detected objects with masks and bounding boxes
+
+**Visualization Topic:** `/segmentedImage` (if enabled)

detectron_ros/detectron_ros/detectron_ros.py

@@ -37,7 +37,7 @@ def __init__(self):
         self.interest_classes = self.declare_parameter("interest_classes", [*range(80)]).value
 
         self.publish_visualization = self.declare_parameter("publish_visualization", True).value
-        visualization_topic = self.declare_parameter("visualization_topic", "/detectron/segmentedImage").value
+        visualization_topic = self.declare_parameter("visualization_topic", "/segmentedImage").value
         self.visualization_pub = self.create_publisher(sensor_msgs.msg.Image, visualization_topic, 1)
 
         self.cv_bridge = CvBridge()