Closed tokuda99 closed 1 week ago
@tokuda99 @amc-nu
Thank you very much for handling this issue, and we are sorry to make things complicated.
Because tar.gz
file this PR refers to is dedicated to the CTI Anvil platform, I guess it probably does not work on RQX-58G. Hence, could you please use v1.4.3 camera driver for RQX-58G platform? This is the latest version that RQX-58G is supported.
@manato, we fixed the issue in the repository. The problem relied on using the latest
tag, causing it to point to another release than the specific one for the rescue.
We used the corresponding ID for the v1.4.3.
We also confirmed the roscube builds successfully.
@manato @drwnz
The new Autoware TensorRT YoloX node uses nested namespaces (autoware::tensorrt_yolox
), which is incompatible with the CUDA compiler in the ROSCUBE. The following change in the code is required.
diff --git a/perception/autoware_tensorrt_yolox/include/autoware/tensorrt_yolox/tensorrt_yolox_node.hpp b/perception/autoware_tensorrt_yolox/include/autoware/tensorrt_yolox/tensorrt_yolox_node.hpp
index 7f622b2dc8..cf7b051b9e 100644
--- a/perception/autoware_tensorrt_yolox/include/autoware/tensorrt_yolox/tensorrt_yolox_node.hpp
+++ b/perception/autoware_tensorrt_yolox/include/autoware/tensorrt_yolox/tensorrt_yolox_node.hpp
@@ -42,8 +42,8 @@
#include <string>
#include <vector>
-namespace autoware::tensorrt_yolox
-{
+namespace autoware {
+namespace tensorrt_yolox {
// cspell: ignore Semseg
using LabelMap = std::map<int, std::string>;
using Label = tier4_perception_msgs::msg::Semantic;
@@ -110,6 +110,6 @@ private:
std::unique_ptr<autoware::universe_utils::DebugPublisher> debug_publisher_;
};
-} // namespace autoware::tensorrt_yolox
-
+} // namespace tensorrt_yolox
+} // namespace autoware
#endif // AUTOWARE__TENSORRT_YOLOX__TENSORRT_YOLOX_NODE_HPP_
diff --git a/perception/autoware_tensorrt_yolox/src/tensorrt_yolox_node.cpp b/perception/autoware_tensorrt_yolox/src/tensorrt_yolox_node.cpp
index f3c7b2552e..ed57c598be 100644
--- a/perception/autoware_tensorrt_yolox/src/tensorrt_yolox_node.cpp
+++ b/perception/autoware_tensorrt_yolox/src/tensorrt_yolox_node.cpp
@@ -24,8 +24,8 @@
#include <utility>
#include <vector>
-namespace autoware::tensorrt_yolox
-{
+namespace autoware {
+namespace tensorrt_yolox {
TrtYoloXNode::TrtYoloXNode(const rclcpp::NodeOptions & node_options)
: Node("tensorrt_yolox", node_options)
{
@@ -281,7 +281,8 @@ void TrtYoloXNode::overlapSegmentByRoi(
.rowRange(roi_y_offset, roi_y_offset + roi_height));
}
-} // namespace autoware::tensorrt_yolox
+} // namespace tensorrt_yolox
+} // namespace autoware
#include "rclcpp_components/register_node_macro.hpp"
RCLCPP_COMPONENTS_REGISTER_NODE(autoware::tensorrt_yolox::TrtYoloXNode)
@amc-nu @tokuda99 @urasakikeisuke
Before merging this PR, would you mind fixing pre-commit.ci
error since it is required to proceed merging process?
@manato fixed the remaining issues. Ansible linting is still failing, but the files are unrelated to this PR.
@amc-nu Thanks for your prompt handling!
Verified using the following command, after a successful installation with ansible and a restart to allow the camera driver to load.
ros2 launch edge_auto_jetson_launch edge_auto_jetson.launch.xml