feat(lidar_transfusion): update TransFusion-L model (autowarefoundati…

…on#7890)

* add num_proposals

Signed-off-by: scepter914 <[email protected]>

* fix config

Signed-off-by: scepter914 <[email protected]>

* update README

Signed-off-by: scepter914 <[email protected]>

---------

Signed-off-by: scepter914 <[email protected]>

perception/lidar_transfusion/README.md

@@ -63,9 +63,9 @@ ros2 launch lidar_transfusion lidar_transfusion.launch.xml log_level:=debug
 
 You can download the onnx format of trained models by clicking on the links below.
 
-- TransFusion: [transfusion.onnx](https://awf.ml.dev.web.auto/perception/models/transfusion/v1/transfusion.onnx)
+- TransFusion: [transfusion.onnx](https://awf.ml.dev.web.auto/perception/models/transfusion/t4xx1_90m/v2/transfusion.onnx)
 
-The model was trained in TIER IV's internal database (~11k lidar frames) for 20 epochs.
+The model was trained in TIER IV's internal database (~11k lidar frames) for 50 epochs.
 
 ### Changelog
 

perception/lidar_transfusion/config/transfusion.param.yaml

@@ -4,8 +4,9 @@
     class_names: ["CAR", "TRUCK", "BUS", "BICYCLE", "PEDESTRIAN"]
     trt_precision: fp16
     voxels_num: [5000, 30000, 60000]  # [min, opt, max]
-    point_cloud_range: [-76.8, -76.8, -3.0, 76.8, 76.8, 5.0]  # [x_min, y_min, z_min, x_max, y_max, z_max]
-    voxel_size: [0.3, 0.3, 8.0]  # [x, y, z]
+    point_cloud_range: [-92.16, -92.16, -3.0, 92.16, 92.16, 7.0]  # [x_min, y_min, z_min, x_max, y_max, z_max]
+    voxel_size: [0.24, 0.24, 10.0]  # [x, y, z]
+    num_proposals: 500
     onnx_path: "$(var model_path)/transfusion.onnx"
     engine_path: "$(var model_path)/transfusion.engine"
     # pre-process params
@@ -17,4 +18,4 @@
     iou_nms_search_distance_2d: 10.0
     iou_nms_threshold: 0.1
     yaw_norm_thresholds: [0.3, 0.3, 0.3, 0.3, 0.0]  # refers to the class_names
-    score_threshold: 0.2
+    score_threshold: 0.1

perception/lidar_transfusion/include/lidar_transfusion/transfusion_config.hpp

@@ -26,8 +26,9 @@ class TransfusionConfig
 public:
   TransfusionConfig(
     const std::vector<int64_t> & voxels_num, const std::vector<double> & point_cloud_range,
-    const std::vector<double> & voxel_size, const float circle_nms_dist_threshold,
-    const std::vector<double> & yaw_norm_thresholds, const float score_threshold)
+    const std::vector<double> & voxel_size, const int num_proposals,
+    const float circle_nms_dist_threshold, const std::vector<double> & yaw_norm_thresholds,
+    const float score_threshold)
   {
     if (voxels_num.size() == 3) {
       max_voxels_ = voxels_num[2];
@@ -61,6 +62,9 @@ class TransfusionConfig
       voxel_y_size_ = static_cast<float>(voxel_size[1]);
       voxel_z_size_ = static_cast<float>(voxel_size[2]);
     }
+    if (num_proposals > 0) {
+      num_proposals_ = num_proposals;
+    }
     if (score_threshold > 0.0) {
       score_threshold_ = score_threshold;
     }
@@ -76,6 +80,9 @@ class TransfusionConfig
     grid_x_size_ = static_cast<std::size_t>((max_x_range_ - min_x_range_) / voxel_x_size_);
     grid_y_size_ = static_cast<std::size_t>((max_y_range_ - min_y_range_) / voxel_y_size_);
     grid_z_size_ = static_cast<std::size_t>((max_z_range_ - min_z_range_) / voxel_z_size_);
+
+    feature_x_size_ = grid_x_size_ / out_size_factor_;
+    feature_y_size_ = grid_y_size_ / out_size_factor_;
   }
 
   ///// INPUT PARAMETERS /////
@@ -107,7 +114,7 @@ class TransfusionConfig
   const std::size_t out_size_factor_{4};
   const std::size_t max_num_points_per_pillar_{points_per_voxel_};
   const std::size_t num_point_values_{4};
-  const std::size_t num_proposals_{200};
+  std::size_t num_proposals_{200};
   // the number of feature maps for pillar scatter
   const std::size_t num_feature_scatter_{pillar_feature_size_};
   // the score threshold for classification

perception/lidar_transfusion/schema/transfusion.schema.json

@@ -61,6 +61,12 @@
           "default": "$(var model_path)/transfusion.engine",
           "pattern": "\\.engine$"
         },
+        "num_proposals": {
+          "type": "integer",
+          "description": "Number of proposals at TransHead.",
+          "default": 500,
+          "minimum": 1
+        },
         "down_sample_factor": {
           "type": "integer",
           "description": "A scale factor of downsampling points",

perception/lidar_transfusion/src/lidar_transfusion_node.cpp

@@ -31,6 +31,7 @@ LidarTransfusionNode::LidarTransfusionNode(const rclcpp::NodeOptions & options)
   const auto point_cloud_range =
     this->declare_parameter<std::vector<double>>("point_cloud_range", descriptor);
   const auto voxel_size = this->declare_parameter<std::vector<double>>("voxel_size", descriptor);
+  const int num_proposals = (this->declare_parameter<int>("num_proposals", descriptor));
   const std::string onnx_path = this->declare_parameter<std::string>("onnx_path", descriptor);
   const std::string engine_path = this->declare_parameter<std::string>("engine_path", descriptor);
 
@@ -73,8 +74,8 @@ LidarTransfusionNode::LidarTransfusionNode(const rclcpp::NodeOptions & options)
   DensificationParam densification_param(
     densification_world_frame_id, densification_num_past_frames);
   TransfusionConfig config(
-    voxels_num, point_cloud_range, voxel_size, circle_nms_dist_threshold, yaw_norm_thresholds,
-    score_threshold);
+    voxels_num, point_cloud_range, voxel_size, num_proposals, circle_nms_dist_threshold,
+    yaw_norm_thresholds, score_threshold);
 
   const auto allow_remapping_by_area_matrix =
     this->declare_parameter<std::vector<int64_t>>("allow_remapping_by_area_matrix", descriptor);