feat(autoware_tensorrt_rtmdet): add tensorrt rtmdet model

build_depends.repos

@@ -58,3 +58,7 @@ repositories:
     type: git
     url: https://github.com/autowarefoundation/ament_cmake.git
     version: feat/faster_ament_libraries_deduplicate
+  universe/external/trt_batched_nms:
+    type: git
+    url: https://github.com/autowarefoundation/trt_batched_nms.git
+    version: main

perception/autoware_tensorrt_rtmdet/CMakeLists.txt

@@ -0,0 +1,105 @@
+cmake_minimum_required(VERSION 3.8)
+project(autoware_tensorrt_rtmdet)
+
+find_package(autoware_tensorrt_common)
+if(NOT ${autoware_tensorrt_common_FOUND})
+  message(WARNING "The autoware_tensorrt_common package is not found. Please check its dependencies.")
+  return()
+endif()
+
+find_package(autoware_cmake REQUIRED)
+autoware_package()
+
+# TODO(amadeuszsz): Remove -Wno-deprecated-declarations once removing implicit quantization
+add_compile_options(-Wno-deprecated-declarations)
+
+find_package(OpenCV REQUIRED)
+
+include(CheckLanguage)
+check_language(CUDA)
+if(CMAKE_CUDA_COMPILER)
+  enable_language(CUDA)
+else()
+  message(WARNING "CUDA is not found. preprocess acceleration using CUDA will not be available.")
+endif()
+
+find_package(OpenMP)
+if(OpenMP_FOUND)
+  set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
+endif()
+
+# tensorrt_rtmdet
+ament_auto_add_library(${PROJECT_NAME} SHARED
+        src/tensorrt_rtmdet.cpp
+)
+
+ament_target_dependencies(${PROJECT_NAME}
+        OpenCV
+)
+
+if(CMAKE_CUDA_COMPILER)
+  # preprocess
+  cuda_add_library(${PROJECT_NAME}_gpu_preprocess
+          SHARED
+          src/preprocess.cu
+  )
+
+  target_include_directories(${PROJECT_NAME}_gpu_preprocess PUBLIC
+          $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
+          $<INSTALL_INTERFACE:include/${PROJECT_NAME}>
+  )
+
+  target_link_libraries(${PROJECT_NAME}
+          ${autoware_tensorrt_common_LIBRARIES}
+          ${PROJECT_NAME}_gpu_preprocess
+  )
+
+else()
+  target_link_libraries(${PROJECT_NAME}
+          ${autoware_tensorrt_common_LIBRARIES}
+  )
+endif()
+
+target_compile_definitions(${PROJECT_NAME} PRIVATE
+        TENSORRT_VERSION_MAJOR=${TENSORRT_VERSION_MAJOR}
+)
+
+# tensorrt_rtmdet_node
+ament_auto_add_library(${PROJECT_NAME}_node SHARED
+        src/tensorrt_rtmdet_node.cpp
+)
+
+ament_target_dependencies(${PROJECT_NAME}_node
+        OpenCV
+)
+
+target_link_libraries(${PROJECT_NAME}_node
+        ${PROJECT_NAME}
+)
+
+target_compile_definitions(${PROJECT_NAME}_node PRIVATE
+        TENSORRT_VERSION_MAJOR=${TENSORRT_VERSION_MAJOR}
+)
+
+rclcpp_components_register_node(${PROJECT_NAME}_node
+        PLUGIN "autoware::tensorrt_rtmdet::TrtRTMDetNode"
+        EXECUTABLE ${PROJECT_NAME}_node_exe
+)
+
+if(BUILD_TESTING)
+  find_package(ament_cmake_ros REQUIRED)
+  find_package(ament_lint_auto REQUIRED)
+  ament_lint_auto_find_test_dependencies()
+
+  ament_add_ros_isolated_gtest(test_rtmdet test/test_rtmdet.cpp)
+  set_tests_properties(test_rtmdet PROPERTIES TIMEOUT 60) # It could take a long time on the first run to create the engine
+  target_link_libraries(test_rtmdet ${PROJECT_NAME} ${PROJECT_NAME}_node)
+  target_include_directories(test_rtmdet PRIVATE include)
+endif()
+
+ament_auto_package(INSTALL_TO_SHARE
+        launch
+        config
+        test/data
+)

perception/autoware_tensorrt_rtmdet/README.md

@@ -0,0 +1,70 @@
+# autoware_tensorrt_rtmdet
+
+## Purpose
+
+RTMDet is a real-time instance segmentation model which can be used for detecting objects like cars, pedestrians,
+bicycles, etc. in a scene. This package provides a ROS 2 interface for RTMDet using TensorRT.
+
+## Inner-workings / Algorithms
+
+## Inputs / Outputs
+
+### Input
+
+| Name       | Type                | Description     |
+| ---------- | ------------------- | --------------- |
+| `in/image` | `sensor_msgs/Image` | The input image |
+
+### Output
+
+| Name              | Type                                                 | Description                                                         |
+| ----------------- | ---------------------------------------------------- | ------------------------------------------------------------------- |
+| `out/objects`     | `tier4_perception_msgs/DetectedObjectsWithFeature`   | The detected objects with 2D bounding boxes and scores              |
+| `out/mask`        | `autoware_internal_perception_msgs/SegmentationMask` | The instance segmentation mask                                      |
+| `out/color_mask`  | `sensor_msgs/Image`                                  | The colorized image of instance segmentation mask for visualization |
+| `out/debug_image` | `sensor_msgs/Image`                                  | The image with 2D bounding boxes for visualization                  |
+
+## Parameters
+
+{{ json_to_markdown("perception/autoware_tensorrt_rtmdet/schema/rtmdet.schema.json") }}
+
+## Assumptions / Known limits
+
+## Onnx model
+
+A sample model is provided in `autoware_data` folder by ansible script on env preparation stage. If you cannot find the
+model, you follow instructions from the
+[link](https://github.com/autowarefoundation/autoware/tree/main/ansible/roles/artifacts) to download the model.
+
+The shared model was trained by open-mmlab using the COCO dataset. For more details,
+see [link](https://github.com/open-mmlab/mmdetection/tree/3.x/configs/rtmdet#instance-segmentation).
+
+### Package acceptable model generation
+
+Users can generate their own model using official RTMDet repository. Please refer to the
+[official repository](https://github.com/open-mmlab/mmdetection/tree/3.x/configs/rtmdet#rtmdet-an-empirical-study-of-designing-real-time-object-detectors)
+
+## Label file
+
+A sample label file (named label.txt) and instance segmentation color map file (name color_map.csv) are also
+downloaded automatically during env preparation process.
+
+These files are used to map the class index to class name and color respectively.
+
+## Reference repositories
+
+- <https://github.com/open-mmlab/mmdetection/tree/3.x/configs/rtmdet#rtmdet-an-empirical-study-of-designing-real-time-object-detectors>
+
+## Citation
+
+```bibtex
+@misc{lyu2022rtmdetempiricalstudydesigning,
+title={RTMDet: An Empirical Study of Designing Real-Time Object Detectors},
+author={Chengqi Lyu and Wenwei Zhang and Haian Huang and Yue Zhou and Yudong Wang and Yanyi Liu and Shilong Zhang and Kai Chen},
+year={2022},
+eprint={2212.07784},
+archivePrefix={arXiv},
+primaryClass={cs.CV},
+url={https://arxiv.org/abs/2212.07784},
+}
+```

perception/autoware_tensorrt_rtmdet/config/rtmdet.param.yaml

@@ -0,0 +1,23 @@
+/**:
+  ros__parameters:
+    is_publish_debug_image: false
+
+    is_publish_color_mask: false
+
+    # cspell: ignore libtrt
+    model_path: "$(var data_path)/tensorrt_rtmdet/$(var model_name).onnx"
+    color_map_path: "$(var data_path)/tensorrt_rtmdet/color_map.csv"
+    plugin_paths: [ "$(find-pkg-prefix trt_batched_nms)/lib/libtrt_batched_nms_plugin.so" ]
+    input_image_mean: [ 103.53, 116.28, 123.675 ]
+    input_image_std_dev: [ 57.375, 57.12, 58.395 ]
+    score_threshold: 0.3
+    nms_threshold: 0.3
+    mask_threshold: 0.7
+    precision: "fp16"
+    dla_core_id: -1
+    quantize_first_layer: false
+    quantize_last_layer: false
+    profile_per_layer: false
+    clip_value: 6.0
+    apply_erosion: false
+    calibration_image_list_path: ""