feat(lidar_centerpoint): priotizing voxel processing by space #1672
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit adds the `priority_map` array to the `preprocess_kernel.hpp` and `centerpoint_trt.hpp` files. The `priority_map` is a new CUDA array that is allocated and initialized in the `PreprocessKernelTest::SetUp()` function. The `priority_map` is used in the `generateBaseFeatures_launch()` function to generate base features for the voxel grid. The `priority_map` is also used in the `centerpoint_trt.hpp` file for initializing and updating the `priority_map_` member variable. Signed-off-by: Taekjin LEE <[email protected]>
…tPriorityMap() This commit modifies the `CenterPointTRT::initPriorityMap()` function to initialize and assign a priority score map. The priority score map is created using the `priority_map` vector, which contains pairs of scores and indices. The map is sorted in descending order based on the scores and then assigned to the `priority_map_` member variable. This change improves the efficiency of the priority map initialization process. feat: initialize and assign priority score map in CenterPointTRT::initPriorityMap() Signed-off-by: Taekjin LEE <[email protected]>
Signed-off-by: Taekjin LEE <[email protected]>
This commit refactors the `CenterPointTRT::initPriorityMap()` function to optimize its performance. The changes include: - Simplifying the calculation of `pos_x` and `pos_y` - Introducing two new variables, `score_a` and `score_b`, to calculate the total score - Adjusting the weights of `score_a` and `score_b` to improve the accuracy of the priority map These optimizations enhance the efficiency and accuracy of the priority map initialization process. Signed-off-by: Taekjin LEE <[email protected]>
Signed-off-by: Taekjin LEE <[email protected]>
Signed-off-by: Taekjin LEE <[email protected]> refactor: optimize CenterPointTRT::initPriorityMap() refactor: optimize CenterPointTRT::initPriorityMap()
technolojin
force-pushed
the
feat/centerpoint/voxel-priority-map
branch
from
4c7e93f to
7e8a3f8
Compare
Signed-off-by: Taekjin LEE <[email protected]>
Signed-off-by: Taekjin LEE <[email protected]>
|
YoshiRi
approved these changes
Dec 11, 2024
|
This PR will be replaced by #1688 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
@TomohitoAndo
こちら、centerpointのmax voxelを超える出力があった際の挙動を調整するPRです。
基本的には今お台場でやっているvoxel sizeの調整がメインで万一漏れた際のセーフティとしてこのPRが活きてくるイメージでいます by Yoshi Ri
Current voxel processing is done by order of X, Y sequentially.
IF the number of filled voxel exceed the ML model limit, the detectable range is limited.
The priority map is set by a scalar field as following.
The following visualization is to show filled voxel distribution. Configurations are as follwing.
Then lower score object will be shown (due to the lower score_threshold), indirectly visualize where the voxel is filled.
Related links
warn the voxel number reaches the limit
feat(autoware_lidar_centerpoint): added a check to notify if we are dropping pillars autowarefoundation/autoware.universe#9488
chore(autoware_lidar_transfusion): added a warning if we are dropping voxels autowarefoundation/autoware.universe#9486
PR for AWF
feat(autoware_lidar_centerpoint): add priority map to preprocess kernel and centerpoint_trt autowarefoundation/autoware.universe#9583
ticket
https://tier4.atlassian.net/browse/RT1-8767
How was this PR tested?
Processing time
left: Before this PR
right: After this PR
Notes for reviewers
None.
Interface changes
None.
Effects on system behavior
None.