lidar_apollo_segmentation_tvm_nodes#
Purpose / Use cases#
An alternative to Euclidean clustering. This node detects and labels foreground obstacles (e.g. cars, motorcycles, pedestrians) from a point cloud, using a neural network.
Design#
See the design of the algorithm in the core (lidar_apollo_segmentation_tvm) package's design documents.
Usage#
lidar_apollo_segmentation_tvm and lidar_apollo_segmentation_tvm_nodes will not work without a neural network.
See the lidar_apollo_segmentation_tvm usage for more information.
Assumptions / Known limits#
The original node from Apollo has a Region Of Interest (ROI) filter. This has the benefit of working with a filtered point cloud that includes only the points inside the ROI (i.e., the drivable road and junction areas) with most of the background obstacles removed (such as buildings and trees around the road region). Not having this filter may negatively impact performance.
Inputs / Outputs / API#
Inputs#
The input are non-ground points as a PointCloud2 message from the sensor_msgs package.
Outputs#
The output is a DetectedObjectsWithFeature.
Parameters#
| Name | Type | Description | Default | Range |
|---|---|---|---|---|
| range | integer | The range of the 2D grid with respect to the origin. | 90 | >0 |
| score_threshold | float | The detection confidence score threshold for filtering out the candidate clusters in the post-processing step. | 0.1 | ≥0.0 ≤1.0 |
| use_intensity_feature | boolean | Enable input channel intensity feature. | false | N/A |
| use_constant_feature | boolean | Enable input channel constant feature. | false | N/A |
| z_offset | float | Vertical translation of the pointcloud before inference. | 0.0 | N/A |
| min_height | float | The minimum height with respect to the origin | -5.0 | N/A |
| max_height | float | The maximum height with respect to the origin. | 5.0 | N/A |
| objectness_thresh | float | The threshold of objectness for filtering out non-object cells in the obstacle clustering step. | 0.5 | ≥0.0 ≤1.0 |
| min_pts_num | integer | In the post-processing step, the candidate clusters with less than min_pts_num points are removed. | 3 | ≥0 |
| height_thresh | float | If it is non-negative, the points that are higher than the predicted object height by height_thresh are filtered out in the post-processing step. | 0.5 | N/A |
| data_path | string | Packages data and artifacts directory path. | $(env HOME)/autoware_data | N/A |
Error detection and handling#
Abort and warn when the input frame can't be converted to base_link.
Security considerations#
Both the input and output are controlled by the same actor, so the following security concerns are out-of-scope:
- Spoofing
- Tampering
Leaking data to another actor would require a flaw in TVM or the host operating system that allows arbitrary memory to be read, a significant security flaw in itself. This is also true for an external actor operating the pipeline early: only the object that initiated the pipeline can run the methods to receive its output.
A Denial-of-Service attack could make the target hardware unusable for other pipelines but would require being able to run code on the CPU, which would already allow a more severe Denial-of-Service attack.
No elevation of privilege is required for this package.
Future extensions / Unimplemented parts#
Related issues#
-
226: Autoware.Auto Neural Networks Inference Architecture Design#