Merge pull request #48 from NVIDIA-ISAAC-ROS/release-3.2

Isaac ROS 3.2
NVIDIA-ISAAC-ROS · Dec 11, 2024 · 06a676b · 06a676b
2 parents 4c47edf + 46da6f1
commit 06a676b
Show file tree

Hide file tree

Showing 45 changed files with 670 additions and 212 deletions.
diff --git a/README.md b/README.md
@@ -63,10 +63,10 @@ This package is powered by [NVIDIA Isaac Transport for ROS (NITROS)](https://dev
 
 ## Performance
 
-| Sample Graph<br/><br/>                                                                                                                                                                                             | Input Size<br/><br/>     | AGX Orin<br/><br/>                                                                                                                                                | Orin NX<br/><br/>                                                                                                                                                | Orin Nano 8GB<br/><br/>                                                                                                                                             | x86_64 w/ RTX 4060 Ti<br/><br/>                                                                                                                                     | x86_64 w/ RTX 4090<br/><br/>                                                                                                                                      |
-|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-| [RT-DETR Object Detection Graph](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/benchmarks/isaac_ros_rtdetr_benchmark/scripts/isaac_ros_rtdetr_graph.py)<br/><br/><br/>SyntheticaDETR<br/><br/> | 720p<br/><br/><br/><br/> | [71.9 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-agx_orin.json)<br/><br/><br/>24 ms @ 30Hz<br/><br/>   | [30.8 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-orin_nx.json)<br/><br/><br/>41 ms @ 30Hz<br/><br/>   | [21.3 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-orin_nano.json)<br/><br/><br/>61 ms @ 30Hz<br/><br/>    | [205 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-nuc_4060ti.json)<br/><br/><br/>8.7 ms @ 30Hz<br/><br/>   | [400 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-x86_4090.json)<br/><br/><br/>6.3 ms @ 30Hz<br/><br/>   |
-| [DetectNet Object Detection Graph](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/benchmarks/isaac_ros_detectnet_benchmark/scripts/isaac_ros_detectnet_graph.py)<br/><br/><br/><br/>            | 544p<br/><br/><br/><br/> | [165 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-agx_orin.json)<br/><br/><br/>20 ms @ 30Hz<br/><br/> | [115 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-orin_nx.json)<br/><br/><br/>26 ms @ 30Hz<br/><br/> | [63.2 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-orin_nano.json)<br/><br/><br/>36 ms @ 30Hz<br/><br/> | [488 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-nuc_4060ti.json)<br/><br/><br/>10 ms @ 30Hz<br/><br/> | [589 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-x86_4090.json)<br/><br/><br/>10 ms @ 30Hz<br/><br/> |
+| Sample Graph<br/><br/>                                                                                                                                                                                             | Input Size<br/><br/>     | AGX Orin<br/><br/>                                                                                                                                                 | Orin NX<br/><br/>                                                                                                                                                 | Orin Nano 8GB<br/><br/>                                                                                                                                             | x86_64 w/ RTX 4090<br/><br/>                                                                                                                                      |
+|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| [RT-DETR Object Detection Graph](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/benchmarks/isaac_ros_rtdetr_benchmark/scripts/isaac_ros_rtdetr_graph.py)<br/><br/><br/>SyntheticaDETR<br/><br/> | 720p<br/><br/><br/><br/> | [56.5 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-agx_orin.json)<br/><br/><br/>30 ms @ 30Hz<br/><br/>    | [33.8 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-orin_nx.json)<br/><br/><br/>39 ms @ 30Hz<br/><br/>    | [24.1 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-orin_nano.json)<br/><br/><br/>53 ms @ 30Hz<br/><br/>    | [490 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_rtdetr_graph-x86-4090.json)<br/><br/><br/>7.1 ms @ 30Hz<br/><br/>   |
+| [DetectNet Object Detection Graph](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/benchmarks/isaac_ros_detectnet_benchmark/scripts/isaac_ros_detectnet_graph.py)<br/><br/><br/><br/>            | 544p<br/><br/><br/><br/> | [70.5 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-agx_orin.json)<br/><br/><br/>26 ms @ 30Hz<br/><br/> | [30.1 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-orin_nx.json)<br/><br/><br/>46 ms @ 30Hz<br/><br/> | [22.9 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-orin_nano.json)<br/><br/><br/>57 ms @ 30Hz<br/><br/> | [254 fps](https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_benchmark/blob/main/results/isaac_ros_detectnet_graph-x86-4090.json)<br/><br/><br/>11 ms @ 30Hz<br/><br/> |
 
 ---
 
@@ -97,4 +97,4 @@ Please visit the [Isaac ROS Documentation](https://nvidia-isaac-ros.github.io/re
 
 ## Latest
 
-Update 2024-09-26: Update for ZED compatibility
+Update 2024-12-10: Update to be compatible with JetPack 6.1
diff --git a/gxf_isaac_detectnet/CMakeLists.txt b/gxf_isaac_detectnet/CMakeLists.txt
@@ -63,4 +63,10 @@ set_target_properties(${PROJECT_NAME} PROPERTIES
 # Install the binary file
 install(TARGETS ${PROJECT_NAME} DESTINATION share/${PROJECT_NAME}/gxf/lib)
 
+
+# Embed versioning information into installed files
+ament_index_get_resource(ISAAC_ROS_COMMON_CMAKE_PATH isaac_ros_common_cmake_path isaac_ros_common)
+include("${ISAAC_ROS_COMMON_CMAKE_PATH}/isaac_ros_common-version-info.cmake")
+generate_version_info(${PROJECT_NAME})
+
 ament_auto_package(INSTALL_TO_SHARE)
diff --git a/gxf_isaac_detectnet/gxf/detectnet/detectnet.cpp b/gxf_isaac_detectnet/gxf/detectnet/detectnet.cpp
@@ -14,11 +14,13 @@
 // limitations under the License.
 //
 // SPDX-License-Identifier: Apache-2.0
+
 #include <string>
+#include <vector>
 
+#include "detectnet/detectnet_decoder.hpp"
 #include "gxf/core/gxf.h"
 #include "gxf/std/extension_factory_helper.hpp"
-#include "detectnet/detectnet_decoder.hpp"
 
 extern "C" {
 

diff --git a/gxf_isaac_detectnet/gxf/detectnet/detectnet_decoder.cpp b/gxf_isaac_detectnet/gxf/detectnet/detectnet_decoder.cpp
@@ -15,18 +15,21 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-#include "detectnet_decoder.hpp"
-#include "detection2_d_array_message.hpp"
-
-#include <string>
 #include <climits>
+#include <memory>
+#include <string>
+#include <vector>
+
+#include "./detection2_d_array_message.hpp"
+#include "./detectnet_decoder.hpp"
+
+#include "cuda.h"
+#include "cuda_runtime.h"
 
+#include "gxf/core/parameter_parser_std.hpp"
 #include "gxf/multimedia/camera.hpp"
 #include "gxf/multimedia/video.hpp"
-#include "gxf/core/parameter_parser_std.hpp"
 #include "gxf/std/timestamp.hpp"
-#include "cuda.h"
-#include "cuda_runtime.h"
 
 
 namespace nvidia
@@ -67,10 +70,10 @@ NvDsInferObjectDetectionInfo GetNewDetectionInfo(
 }
 
 void FillMessage(
-  Detection2DParts &message_parts,
-  const std::vector<NvDsInferObjectDetectionInfo> &detection_info_vector,
+  Detection2DParts& message_parts,
+  const std::vector<NvDsInferObjectDetectionInfo>& detection_info_vector,
   gxf::Handle<nvidia::gxf::Timestamp> tensorlist_timestamp,
-  size_t num_detections, const std::vector<std::string> &label_list)
+  size_t num_detections, const std::vector<std::string>& label_list)
 {
   for (uint32_t i = 0; i < num_detections; i++) {
     NvDsInferObjectDetectionInfo detection_info = detection_info_vector[i];
@@ -89,7 +92,7 @@ void FillMessage(
   }
   *(message_parts.timestamp) = *tensorlist_timestamp;
 }
-} // anonymous namespace
+}  // anonymous namespace
 
 
 gxf_result_t DetectnetDecoder::registerInterface(gxf::Registrar * registrar) noexcept
@@ -106,8 +109,8 @@ gxf_result_t DetectnetDecoder::registerInterface(gxf::Registrar * registrar) noe
 
   result &= registrar->parameter(
     label_list_, "label_list", "List of network labels",
-    "List of labels corresponding to the int labels received from the tensors", {"person", "bag",
-      "face"});
+    "List of labels corresponding to the int labels received from the tensors",
+    {"person", "bag", "face"});
 
   result &= registrar->parameter(
     enable_confidence_threshold_, "enable_confidence_threshold", "Enable Confidence Threshold",
@@ -131,27 +134,27 @@ gxf_result_t DetectnetDecoder::registerInterface(gxf::Registrar * registrar) noe
 
   result &= registrar->parameter(
     dbscan_confidence_threshold_, "dbscan_confidence_threshold", "Dbscan Confidence Threshold",
-    "Minimum score in a cluster for the cluster to be considered an object \
-     during grouping. Different clustering may cause the algorithm \
-     to use different scores.",
+    "Minimum score in a cluster for the cluster to be considered an object "
+    "during grouping. Different clustering may cause the algorithm "
+    "to use different scores.",
     0.6);
 
   result &= registrar->parameter(
     dbscan_eps_, "dbscan_eps", "Dbscan Epsilon",
-    "Holds the epsilon to control merging of overlapping boxes. \
-    Refer to OpenCV groupRectangles and DBSCAN documentation for more information on epsilon. ",
+    "Holds the epsilon to control merging of overlapping boxes. "
+    "Refer to OpenCV groupRectangles and DBSCAN documentation for more information on epsilon. ",
     0.01);
 
   result &= registrar->parameter(
     dbscan_min_boxes_, "dbscan_min_boxes", "Dbscan Minimum Boxes",
-    "Holds the minimum number of boxes in a cluster to be considered \
-     an object during grouping using DBSCAN",
+    "Holds the minimum number of boxes in a cluster to be considered "
+    "an object during grouping using DBSCAN",
     1);
 
   result &= registrar->parameter(
     dbscan_enable_athr_filter_, "dbscan_enable_athr_filter", "Dbscan Enable Athr Filter",
-    "true enables the area-to-hit ratio (ATHR) filter. \
-     The ATHR is calculated as: ATHR = sqrt(clusterArea) / nObjectsInCluster.",
+    "true enables the area-to-hit ratio (ATHR) filter. "
+    "The ATHR is calculated as: ATHR = sqrt(clusterArea) / nObjectsInCluster.",
     0);
 
   result &= registrar->parameter(
@@ -196,7 +199,6 @@ gxf_result_t DetectnetDecoder::start() noexcept
 
 gxf_result_t DetectnetDecoder::tick() noexcept
 {
-
   gxf::Expected<void> result;
 
   // Receive disparity image and left/right camera info
@@ -272,9 +274,10 @@ gxf_result_t DetectnetDecoder::tick() noexcept
     return GXF_FAILURE;
   }
 
-  float bbox_tensor_arr[bbox_tensor->size() / sizeof(float)]; // since data in tensor is kFloat32
+  // data in tensor is kFloat32
+  std::vector<float> bbox_tensor_arr(bbox_tensor->size() / sizeof(float));
   const cudaError_t cuda_error_bbox_tensor = cudaMemcpy(
-    &bbox_tensor_arr, bbox_tensor->pointer(),
+    bbox_tensor_arr.data(), bbox_tensor->pointer(),
     bbox_tensor->size(), cudaMemcpyDeviceToHost);
   if (cuda_error_bbox_tensor != cudaSuccess) {
     GXF_LOG_ERROR("Error while copying kernel: %s", cudaGetErrorString(cuda_error_bbox_tensor));
@@ -315,8 +318,8 @@ gxf_result_t DetectnetDecoder::tick() noexcept
         float coverage = cov_tensor_arr[cov_pos];
 
         // Center of the grid in pixels
-        float grid_center_y = (row + bounding_box_offset_ ) * kStride;
-        float grid_center_x = (col + bounding_box_offset_ ) * kStride;
+        float grid_center_y = (row + bounding_box_offset_) * kStride;
+        float grid_center_x = (col + bounding_box_offset_) * kStride;
 
         // Get each element of the bounding box
         float bbox[kBoundingBoxParams];
@@ -342,7 +345,8 @@ gxf_result_t DetectnetDecoder::tick() noexcept
         // check if object_class is out of range for label_list_
         if (static_cast<size_t>(object_class) >= label_list_.get().size()) {
           GXF_LOG_ERROR(
-            "[DetectNet Decoder] object_class %i is out of range for provided label_list_ of size %lu", object_class,
+            "[DetectNet Decoder] object_class %i is out of range for provided "
+            "label_list_ of size %lu", object_class,
             label_list_.get().size());
           return GXF_FAILURE;
         }
@@ -360,7 +364,7 @@ gxf_result_t DetectnetDecoder::tick() noexcept
 
   size_t num_detections = detection_info_vector.size();
   if (enable_dbscan_clustering_) {
-    NvDsInferObjectDetectionInfo * detection_info_pointer = &detection_info_vector[0];
+    NvDsInferObjectDetectionInfo* detection_info_pointer = &detection_info_vector[0];
     NvDsInferDBScanHandle dbscan_hdl = NvDsInferDBScanCreate();
     if (dbscan_clustering_algorithm_ == kDbscanCluster) {
       NvDsInferDBScanCluster(dbscan_hdl, &params_, detection_info_pointer, &num_detections);
@@ -386,7 +390,6 @@ gxf_result_t DetectnetDecoder::tick() noexcept
           num_detections, label_list_);
         return detections_transmitter_->publish(message_parts.message);
       }));
-
 }
 }  // namespace isaac_ros
 }  // namespace nvidia
diff --git a/gxf_isaac_detectnet/gxf/detectnet/detectnet_decoder.hpp b/gxf_isaac_detectnet/gxf/detectnet/detectnet_decoder.hpp
@@ -18,15 +18,19 @@
 #ifndef NVIDIA_ISAAC_ROS_EXTENSIONS_DETECTNET_DECODER_HPP_
 #define NVIDIA_ISAAC_ROS_EXTENSIONS_DETECTNET_DECODER_HPP_
 
+#include <string>
+#include <vector>
+
+#include "./detection2_d_array_message.hpp"
+#include "deepstream_utils/nvdsinferutils/dbscan/nvdsinfer_dbscan.hpp"
+
 #include "gxf/core/entity.hpp"
 #include "gxf/core/gxf.h"
 #include "gxf/core/parameter.hpp"
-#include "gxf/std/codelet.hpp"
 #include "gxf/core/parameter_parser_std.hpp"
+#include "gxf/std/codelet.hpp"
 #include "gxf/std/receiver.hpp"
 #include "gxf/std/transmitter.hpp"
-#include "detection2_d_array_message.hpp"
-#include "deepstream_utils/nvdsinferutils/dbscan/nvdsinfer_dbscan.hpp"
 
 
 namespace nvidia

diff --git a/gxf_isaac_detectnet/package.xml b/gxf_isaac_detectnet/package.xml
@@ -21,7 +21,7 @@ SPDX-License-Identifier: Apache-2.0
 <?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
 <package format="3">
   <name>gxf_isaac_detectnet</name>
-  <version>3.1.0</version>
+  <version>3.2.0</version>
   <description>Detectnet GXF extension.</description>
 
   <maintainer email="[email protected]">Isaac ROS Maintainers</maintainer>

diff --git a/isaac_ros_detectnet/CMakeLists.txt b/isaac_ros_detectnet/CMakeLists.txt
@@ -47,7 +47,7 @@ if(BUILD_TESTING)
   endif()
 
   find_package(launch_testing_ament_cmake REQUIRED)
-  add_launch_test(test/isaac_ros_detectnet_pol_test.py TIMEOUT "600")
+  add_launch_test(test/isaac_ros_detectnet_pol_test.py TIMEOUT "900")
 endif()
 
 # Visualizer python scripts
@@ -63,4 +63,10 @@ install(DIRECTORY
   DESTINATION share/${PROJECT_NAME}
 )
 
+
+# Embed versioning information into installed files
+ament_index_get_resource(ISAAC_ROS_COMMON_CMAKE_PATH isaac_ros_common_cmake_path isaac_ros_common)
+include("${ISAAC_ROS_COMMON_CMAKE_PATH}/isaac_ros_common-version-info.cmake")
+generate_version_info(${PROJECT_NAME})
+
 ament_auto_package(INSTALL_TO_SHARE config launch)
diff --git a/isaac_ros_detectnet/config/hawk_config.pbtxt b/isaac_ros_detectnet/config/hawk_config.pbtxt
@@ -3,20 +3,20 @@ platform: "tensorrt_plan"
 max_batch_size: 16
 input [
   {
-    name: "input_1"
+    name: "input_1:0"
     data_type: TYPE_FP32
     format: FORMAT_NCHW
     dims: [ 3, 1200, 1920 ]
   }
 ]
 output [
   {
-        name: "output_bbox/BiasAdd"
+        name: "output_bbox/BiasAdd:0"
         data_type: TYPE_FP32
         dims: [ 12, 75, 120]
     },
     {
-        name: "output_cov/Sigmoid"
+        name: "output_cov/Sigmoid:0"
         data_type: TYPE_FP32
         dims: [ 3, 75, 120]
     }

diff --git a/isaac_ros_detectnet/config/isaac_sim_config.pbtxt b/isaac_ros_detectnet/config/isaac_sim_config.pbtxt
@@ -3,20 +3,20 @@ platform: "tensorrt_plan"
 max_batch_size: 16
 input [
   {
-    name: "input_1"
+    name: "input_1:0"
     data_type: TYPE_FP32
     format: FORMAT_NCHW
     dims: [ 3, 720, 1280 ]
   }
 ]
 output [
   {
-        name: "output_bbox/BiasAdd"
+        name: "output_bbox/BiasAdd:0"
         data_type: TYPE_FP32
         dims: [ 12, 45, 80]
     },
     {
-        name: "output_cov/Sigmoid"
+        name: "output_cov/Sigmoid:0"
         data_type: TYPE_FP32
         dims: [ 3, 45, 80]
     }

diff --git a/isaac_ros_detectnet/config/peoplenet_config.pbtxt b/isaac_ros_detectnet/config/peoplenet_config.pbtxt
@@ -3,22 +3,22 @@ platform: "tensorrt_plan"
 max_batch_size: 16
 input [
   {
-    name: "input_1"
+    name: "input_1:0"
     data_type: TYPE_FP32
     format: FORMAT_NCHW
-    dims: [ 3, 544, 960 ]
+    dims: [ 3, 544, 960]
   }
 ]
 output [
   {
-        name: "output_bbox/BiasAdd"
+        name: "output_bbox/BiasAdd:0"
         data_type: TYPE_FP32
-        dims: [ 12, 34, 60 ]
+        dims: [ 12, 34, 60]
     },
     {
-        name: "output_cov/Sigmoid"
+        name: "output_cov/Sigmoid:0"
         data_type: TYPE_FP32
-        dims: [ 3, 34, 60 ]
+        dims: [ 3, 34, 60]
     }
 ]
 dynamic_batching { }

diff --git a/isaac_ros_detectnet/config/quickstart_config.pbtxt b/isaac_ros_detectnet/config/quickstart_config.pbtxt
@@ -3,22 +3,22 @@ platform: "tensorrt_plan"
 max_batch_size: 16
 input [
   {
-    name: "input_1"
+    name: "input_1:0"
     data_type: TYPE_FP32
     format: FORMAT_NCHW
-    dims: [ 3, 632, 1200 ]
+    dims: [ 3, 544, 960]
   }
 ]
 output [
   {
-        name: "output_bbox/BiasAdd"
+        name: "output_bbox/BiasAdd:0"
         data_type: TYPE_FP32
-        dims: [ 12, 40, 75]
+        dims: [ 12, 34, 60]
     },
     {
-        name: "output_cov/Sigmoid"
+        name: "output_cov/Sigmoid:0"
         data_type: TYPE_FP32
-        dims: [ 3, 40, 75]
+        dims: [ 3, 34, 60]
     }
 ]
 dynamic_batching { }

diff --git a/isaac_ros_detectnet/config/realsense_config.pbtxt b/isaac_ros_detectnet/config/realsense_config.pbtxt
@@ -3,20 +3,20 @@ platform: "tensorrt_plan"
 max_batch_size: 16
 input [
   {
-    name: "input_1"
+    name: "input_1:0"
     data_type: TYPE_FP32
     format: FORMAT_NCHW
     dims: [ 3, 480, 640 ]
   }
 ]
 output [
   {
-        name: "output_bbox/BiasAdd"
+        name: "output_bbox/BiasAdd:0"
         data_type: TYPE_FP32
         dims: [ 12, 30, 40]
     },
     {
-        name: "output_cov/Sigmoid"
+        name: "output_cov/Sigmoid:0"
         data_type: TYPE_FP32
         dims: [ 3, 30, 40]
     }