Skip to content

feat(deployment): centerpoint deployment integration #181

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
vividf wants to merge 2 commits into
base: feat/new_deployment_and_evaluation_pipeline
Choose a base branch
from
Open

feat(deployment): centerpoint deployment integration #181

vividf wants to merge 2 commits into from

Conversation

@vividf
Copy link
Collaborator

@vividf vividf commented Feb 2, 2026 •
edited
Loading

Summary

Integrates CenterPoint into the unified deployment framework, enabling deployment and evaluation of ONNX and TensorRT models.

Note, this PR include changes in #180

Changes

  • Integrated CenterPoint with deployment framework:
    • Moved deployment code from projects/CenterPoint to deployment/projects/centerpoint
    • Implemented component-based export pipeline for ONNX and TensorRT
    • Added runtime inference support with PyTorch, ONNX Runtime, and TensorRT backends
  • Deployment capabilities:
    • Export CenterPoint models to ONNX format
    • Export CenterPoint models to TensorRT engines
    • Component-based architecture (voxel encoder, backbone+head) for flexible deployment
  • Evaluation capabilities:
    • Evaluate ONNX models using ONNX Runtime
    • Evaluate TensorRT engines
    • Integrated metrics evaluation with deployment pipeline
  • Updated CLI: Replaced old deploy.py script with new unified CLI (deployment.cli.main)
  • Added Docker support: Created Dockerfile for deployment environment with TensorRT dependencies
  • Updated documentation: Added deployment and evaluation instructions in README

Migration Notes

  • Old deployment script (projects/CenterPoint/scripts/deploy.py) is removed
  • Use new CLI: python -m deployment.cli.main centerpoint <deploy_config> <model_config>
  • ONNX model variants are now registered via deployment.projects.centerpoint.onnx_models

How to run

python -m deployment.cli.main centerpoint   deployment/projects/centerpoint/config/deploy_config.py   projects/CenterPoint/configs/t4dataset/Centerpoint/second_secfpn_4xb16_121m_j6gen2_base_t4metric_v2.py   --rot-y-axis-reference

Exported ONNX (Same)

Voxel Encoder
Screenshot from 2025-11-28 16-51-47

Backbone Head
Screenshot from 2025-11-28 16-50-46

Test with Deployment pipeline

 python -m deployment.cli.main centerpoint   deployment/projects/centerpoint/config/deploy_config.py   projects/CenterPoint/configs/t4dataset/Centerpoint/second_secfpn_4xb16_121m_j6gen2_base_amp_t4metric_v2.py   --rot-y-axis-reference

Example: TENSORRT

Frame:
Total Num: 1
Skipped Frames: []
Skipped Frames Count: 0

Ground Truth Num: 37

mAP: 0.7897, mAPH: 0.7422 (Center Distance BEV)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 32 18 0.7749 0.7634 0.8125 0.9286 0.7222 0.307514
1.00 32 18 0.8136 0.8020 0.8125 0.9286 0.7222 0.307514
2.00 32 18 0.8342 0.8228 0.8750 1.0000 0.7778 0.307514
4.00 32 18 0.8342 0.8228 0.8750 1.0000 0.7778 0.307514

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 16 8 0.5778 0.5778 0.7692 1.0000 0.6250 0.597714
1.00 16 8 0.7145 0.7111 0.8571 1.0000 0.7500 0.372688
2.00 16 8 0.7145 0.7111 0.8571 1.0000 0.7500 0.372688
4.00 16 8 0.7653 0.7607 0.8571 1.0000 0.7500 0.372688

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974
1.00 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974
2.00 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974
4.00 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362
1.00 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362
2.00 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362
4.00 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 19 9 0.4206 0.2632 0.5882 0.6250 0.5556 0.265326
1.00 19 9 0.4206 0.2632 0.5882 0.6250 0.5556 0.265326
2.00 19 9 0.4206 0.2632 0.5882 0.6250 0.5556 0.265326
4.00 19 9 0.5611 0.3503 0.7059 0.7500 0.6667 0.265326

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 32 18 0.50/1.00/2.00/4.00 0.8143 0.7749 / 0.8136 / 0.8342 / 0.8342 0.803 0.7634 / 0.8020 / 0.8228 / 0.8228
truck 16 8 0.50/1.00/2.00/4.00 0.6930 0.5778 / 0.7145 / 0.7145 / 0.7653 0.690 0.5778 / 0.7111 / 0.7111 / 0.7607
bus 3 1 0.50/1.00/2.00/4.00 0.9918 0.9918 / 0.9918 / 0.9918 / 0.9918 0.989 0.9889 / 0.9889 / 0.9889 / 0.9889
bicycle 2 1 0.50/1.00/2.00/4.00 0.9938 0.9938 / 0.9938 / 0.9938 / 0.9938 0.944 0.9444 / 0.9444 / 0.9444 / 0.9444
pedestrian 19 9 0.50/1.00/2.00/4.00 0.4557 0.4206 / 0.4206 / 0.4206 / 0.5611 0.285 0.2632 / 0.2632 / 0.2632 / 0.3503

mAP: 0.8050, mAPH: 0.7548 (Plane Distance)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 32 18 0.8342 0.8228 0.8750 1.0000 0.7778 0.307514
4.00 32 18 0.8342 0.8228 0.8750 1.0000 0.7778 0.307514

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 16 8 0.7145 0.7111 0.8571 1.0000 0.7500 0.372688
4.00 16 8 0.7145 0.7111 0.8571 1.0000 0.7500 0.372688

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974
4.00 3 1 0.9918 0.9889 1.0000 1.0000 1.0000 0.898974

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362
4.00 2 1 0.9938 0.9444 1.0000 1.0000 1.0000 0.710362

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 19 9 0.4206 0.2632 0.5882 0.6250 0.5556 0.265326
4.00 19 9 0.5611 0.3503 0.7059 0.7500 0.6667 0.265326

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 32 18 2.00/4.00 0.8342 0.8342 / 0.8342 0.823 0.8228 / 0.8228
truck 16 8 2.00/4.00 0.7145 0.7145 / 0.7145 0.711 0.7111 / 0.7111
bus 3 1 2.00/4.00 0.9918 0.9918 / 0.9918 0.989 0.9889 / 0.9889
bicycle 2 1 2.00/4.00 0.9938 0.9938 / 0.9938 0.944 0.9444 / 0.9444
pedestrian 19 9 2.00/4.00 0.4908 0.4206 / 0.5611 0.307 0.2632 / 0.3503

Latency Statistics:
Mean: 841.05 ms
Std: 0.00 ms
Min: 841.05 ms
Max: 841.05 ms
Median: 841.05 ms

Stage-wise Latency Breakdown:
Preprocessing : 515.14 ± 0.00 ms
Model : 324.29 ± 0.00 ms
Voxel Encoder : 3.14 ± 0.00 ms
Middle Encoder : 1.05 ± 0.00 ms
Backbone Head : 133.01 ± 0.00 ms
Postprocessing : 1.60 ± 0.00 ms

vividf added 2 commits February 2, 2026 19:06
@vividf
feat: refactor config
7d0eabe 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
feat: integrate centerpoint to deployment framework
37fae31 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf vividf changed the title Feat/centerpoint deployment integration feat(deployment): centerpoint deployment integration Feb 2, 2026
@vividf vividf mentioned this pull request Feb 2, 2026
Closed
@vividf vividf requested review from KSeangTan and yamsam February 2, 2026 16:33
@vividf vividf self-assigned this Feb 2, 2026
@vividf vividf marked this pull request as ready for review February 3, 2026 04:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@KSeangTan KSeangTan Awaiting requested review from KSeangTan

@yamsam yamsam Awaiting requested review from yamsam

Assignees

@vividf vividf

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vividf