Skip to content

feat(centerpoint): integrate CenterPoint into unified deployment pipeline #161

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
Closed
vividf wants to merge 16 commits into from
Closed

feat(centerpoint): integrate CenterPoint into unified deployment pipeline #161

vividf wants to merge 16 commits into from

Conversation

@vividf
Copy link
Collaborator

@vividf vividf commented Dec 25, 2025 •
edited
Loading

Summary

  • Adds a CenterPoint project bundle under deployment/projects/centerpoint/ and hooks it into the unified deployment CLI via the project registry.
  • Supports PyTorch / ONNXRuntime / TensorRT export + evaluation with a componentized (multi-file) ONNX → per-component TRT engine flow.
  • Removes legacy CenterPoint deployment entrypoints in projects/CenterPoint/* in favor of the unified framework.

Change points

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

Evaluation result (Same for Deployment and Test)

Test with test.py

python tools/detection3d/test.py projects/CenterPoint/configs/t4dataset/Centerpoint/second_secfpn_4xb16_121m_j6gen2_base_t4metric_v2.py work_dirs/centerpoint/best_checkpoint.pth

Frame:
Total Num: 19
Skipped Frames: []
Skipped Frames Count: 0

Ground Truth Num: 650

mAP: 0.5860, mAPH: 0.5621 (Center Distance BEV)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 583 337 0.7919 0.7744 0.8207 0.9007 0.7537 0.331500
1.00 583 337 0.8425 0.8242 0.8507 0.9266 0.7864 0.321666
2.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666
4.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 278 152 0.5563 0.5457 0.7237 0.8857 0.6118 0.358953
1.00 278 152 0.8317 0.8122 0.8671 0.9254 0.8158 0.225477
2.00 278 152 0.8459 0.8276 0.8741 0.9328 0.8224 0.225477
4.00 278 152 0.8675 0.8476 0.8741 0.9328 0.8224 0.225477

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
1.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 1 19 0.0000 0.0000 nan nan nan nan
1.00 1 19 0.0000 0.0000 nan nan nan nan
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 328 123 0.3030 0.2310 0.5123 0.4506 0.5935 0.222946
1.00 328 123 0.3030 0.2310 0.5123 0.4506 0.5935 0.222946
2.00 328 123 0.3166 0.2417 0.5193 0.4568 0.6016 0.222946
4.00 328 123 0.3489 0.2668 0.5422 0.4306 0.7317 0.180424

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 583 337 0.50/1.00/2.00/4.00 0.8389 0.7919 / 0.8425 / 0.8607 / 0.8607 0.821 0.7744 / 0.8242 / 0.8421 / 0.8421
truck 278 152 0.50/1.00/2.00/4.00 0.7754 0.5563 / 0.8317 / 0.8459 / 0.8675 0.758 0.5457 / 0.8122 / 0.8276 / 0.8476
bus 23 19 0.50/1.00/2.00/4.00 0.9979 0.9979 / 0.9979 / 0.9979 / 0.9979 0.989 0.9889 / 0.9889 / 0.9889 / 0.9889
bicycle 1 19 0.50/1.00/2.00/4.00 0.0000 0.0000 / 0.0000 / 0.0000 / 0.0000 0.000 0.0000 / 0.0000 / 0.0000 / 0.0000
pedestrian 328 123 0.50/1.00/2.00/4.00 0.3179 0.3030 / 0.3030 / 0.3166 / 0.3489 0.243 0.2310 / 0.2310 / 0.2417 / 0.2668

mAP: 0.6076, mAPH: 0.5831 (Plane Distance)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 583 337 0.8602 0.8422 0.8957 0.9755 0.8279 0.321666
4.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 278 152 0.8410 0.8281 0.8671 0.9254 0.8158 0.225477
4.00 278 152 0.8526 0.8329 0.8741 0.9328 0.8224 0.225477

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 328 123 0.3166 0.2417 0.5193 0.4568 0.6016 0.222946
4.00 328 123 0.3490 0.2664 0.5422 0.4306 0.7317 0.180424

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 583 337 2.00/4.00 0.8604 0.8602 / 0.8607 0.842 0.8422 / 0.8421
truck 278 152 2.00/4.00 0.8468 0.8410 / 0.8526 0.831 0.8281 / 0.8329
bus 23 19 2.00/4.00 0.9979 0.9979 / 0.9979 0.989 0.9889 / 0.9889
bicycle 1 19 2.00/4.00 0.0000 0.0000 / 0.0000 0.000 0.0000 / 0.0000
pedestrian 328 123 2.00/4.00 0.3328 0.3166 / 0.3490 0.254 0.2417 / 0.2664

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_t4metric_v2.py   --rot-y-axis-reference

PYTORCH Results:

mAP: 0.5860, mAPH: 0.5621 (Center Distance BEV)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 583 337 0.7919 0.7744 0.8207 0.9007 0.7537 0.331500
1.00 583 337 0.8425 0.8242 0.8507 0.9266 0.7864 0.321666
2.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666
4.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 278 152 0.5563 0.5457 0.7237 0.8857 0.6118 0.358953
1.00 278 152 0.8317 0.8122 0.8671 0.9254 0.8158 0.225477
2.00 278 152 0.8459 0.8276 0.8741 0.9328 0.8224 0.225477
4.00 278 152 0.8675 0.8476 0.8741 0.9328 0.8224 0.225477

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
1.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 1 19 0.0000 0.0000 nan nan nan nan
1.00 1 19 0.0000 0.0000 nan nan nan nan
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 328 123 0.3030 0.2310 0.5123 0.4506 0.5935 0.222946
1.00 328 123 0.3030 0.2310 0.5123 0.4506 0.5935 0.222946
2.00 328 123 0.3166 0.2417 0.5193 0.4568 0.6016 0.222946
4.00 328 123 0.3489 0.2668 0.5422 0.4306 0.7317 0.180424

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 583 337 0.50/1.00/2.00/4.00 0.8389 0.7919 / 0.8425 / 0.8607 / 0.8607 0.821 0.7744 / 0.8242 / 0.8421 / 0.8421
truck 278 152 0.50/1.00/2.00/4.00 0.7754 0.5563 / 0.8317 / 0.8459 / 0.8675 0.758 0.5457 / 0.8122 / 0.8276 / 0.8476
bus 23 19 0.50/1.00/2.00/4.00 0.9979 0.9979 / 0.9979 / 0.9979 / 0.9979 0.989 0.9889 / 0.9889 / 0.9889 / 0.9889
bicycle 1 19 0.50/1.00/2.00/4.00 0.0000 0.0000 / 0.0000 / 0.0000 / 0.0000 0.000 0.0000 / 0.0000 / 0.0000 / 0.0000
pedestrian 328 123 0.50/1.00/2.00/4.00 0.3179 0.3030 / 0.3030 / 0.3166 / 0.3489 0.243 0.2310 / 0.2310 / 0.2417 / 0.2668

mAP: 0.6076, mAPH: 0.5831 (Plane Distance)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 583 337 0.8602 0.8422 0.8957 0.9755 0.8279 0.321666
4.00 583 337 0.8607 0.8421 0.8957 0.9755 0.8279 0.321666

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 278 152 0.8410 0.8281 0.8671 0.9254 0.8158 0.225477
4.00 278 152 0.8526 0.8329 0.8741 0.9328 0.8224 0.225477

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849640

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 328 123 0.3166 0.2417 0.5193 0.4568 0.6016 0.222946
4.00 328 123 0.3490 0.2664 0.5422 0.4306 0.7317 0.180424

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 583 337 2.00/4.00 0.8604 0.8602 / 0.8607 0.842 0.8422 / 0.8421
truck 278 152 2.00/4.00 0.8468 0.8410 / 0.8526 0.831 0.8281 / 0.8329
bus 23 19 2.00/4.00 0.9979 0.9979 / 0.9979 0.989 0.9889 / 0.9889
bicycle 1 19 2.00/4.00 0.0000 0.0000 / 0.0000 0.000 0.0000 / 0.0000
pedestrian 328 123 2.00/4.00 0.3328 0.3166 / 0.3490 0.254 0.2417 / 0.2664

ONNX Results:

mAP: 0.5860, mAPH: 0.5619 (Center Distance BEV)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 580 337 0.7919 0.7745 0.8207 0.9007 0.7537 0.332124
1.00 580 337 0.8425 0.8242 0.8507 0.9266 0.7864 0.325795
2.00 580 337 0.8606 0.8414 0.8957 0.9755 0.8279 0.325795
4.00 580 337 0.8606 0.8414 0.8957 0.9755 0.8279 0.325795

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 278 152 0.5516 0.5411 0.7237 0.8857 0.6118 0.358248
1.00 278 152 0.8307 0.8106 0.8662 0.9318 0.8092 0.227877
2.00 278 152 0.8454 0.8265 0.8732 0.9394 0.8158 0.227877
4.00 278 152 0.8672 0.8466 0.8732 0.9394 0.8158 0.227877

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546
1.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 1 19 0.0000 0.0000 nan nan nan nan
1.00 1 19 0.0000 0.0000 nan nan nan nan
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 330 123 0.3050 0.2325 0.5123 0.4506 0.5935 0.221067
1.00 330 123 0.3050 0.2325 0.5123 0.4506 0.5935 0.221067
2.00 330 123 0.3185 0.2431 0.5193 0.4568 0.6016 0.221067
4.00 330 123 0.3506 0.2683 0.5410 0.4320 0.7236 0.183189

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 580 337 0.50/1.00/2.00/4.00 0.8389 0.7919 / 0.8425 / 0.8606 / 0.8606 0.820 0.7745 / 0.8242 / 0.8414 / 0.8414
truck 278 152 0.50/1.00/2.00/4.00 0.7737 0.5516 / 0.8307 / 0.8454 / 0.8672 0.756 0.5411 / 0.8106 / 0.8265 / 0.8466
bus 23 19 0.50/1.00/2.00/4.00 0.9979 0.9979 / 0.9979 / 0.9979 / 0.9979 0.989 0.9889 / 0.9889 / 0.9889 / 0.9889
bicycle 1 19 0.50/1.00/2.00/4.00 0.0000 0.0000 / 0.0000 / 0.0000 / 0.0000 0.000 0.0000 / 0.0000 / 0.0000 / 0.0000
pedestrian 330 123 0.50/1.00/2.00/4.00 0.3198 0.3050 / 0.3050 / 0.3185 / 0.3506 0.244 0.2325 / 0.2325 / 0.2431 / 0.2683

mAP: 0.6072, mAPH: 0.5830 (Plane Distance)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 580 337 0.8541 0.8409 0.8957 0.9755 0.8279 0.325795
4.00 580 337 0.8606 0.8414 0.8957 0.9755 0.8279 0.325795

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 278 152 0.8402 0.8278 0.8662 0.9318 0.8092 0.227877
4.00 278 152 0.8523 0.8319 0.8732 0.9394 0.8158 0.227877

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849546

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 1 19 0.0000 0.0000 nan nan nan nan
4.00 1 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 330 123 0.3185 0.2431 0.5193 0.4568 0.6016 0.221067
4.00 330 123 0.3506 0.2676 0.5433 0.4292 0.7398 0.179694

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 580 337 2.00/4.00 0.8574 0.8541 / 0.8606 0.841 0.8409 / 0.8414
truck 278 152 2.00/4.00 0.8463 0.8402 / 0.8523 0.830 0.8278 / 0.8319
bus 23 19 2.00/4.00 0.9979 0.9979 / 0.9979 0.989 0.9889 / 0.9889
bicycle 1 19 2.00/4.00 0.0000 0.0000 / 0.0000 0.000 0.0000 / 0.0000
pedestrian 330 123 2.00/4.00 0.3345 0.3185 / 0.3506 0.255 0.2431 / 0.2676

TENSORRT Results:

mAP: 0.5866, mAPH: 0.5629 (Center Distance BEV)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 586 337 0.7896 0.7718 0.8148 0.8908 0.7507 0.330931
1.00 586 337 0.8421 0.8235 0.8480 0.9201 0.7864 0.323492
2.00 586 337 0.8604 0.8412 0.8928 0.9688 0.8279 0.323492
4.00 586 337 0.8604 0.8412 0.8928 0.9688 0.8279 0.323492

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 282 152 0.5480 0.5375 0.7160 0.8762 0.6053 0.361066
1.00 282 152 0.8328 0.8190 0.8683 0.9457 0.8026 0.244491
2.00 282 152 0.8509 0.8329 0.8819 0.9338 0.8355 0.229696
4.00 282 152 0.8707 0.8520 0.8819 0.9338 0.8355 0.229696

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514
1.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 2 19 0.0000 0.0000 nan nan nan nan
1.00 2 19 0.0000 0.0000 nan nan nan nan
2.00 2 19 0.0000 0.0000 nan nan nan nan
4.00 2 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
0.50 324 123 0.3065 0.2339 0.5123 0.4506 0.5935 0.225948
1.00 324 123 0.3065 0.2339 0.5123 0.4506 0.5935 0.225948
2.00 324 123 0.3201 0.2452 0.5193 0.4568 0.6016 0.225948
4.00 324 123 0.3529 0.2698 0.5488 0.4390 0.7317 0.181541

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 586 337 0.50/1.00/2.00/4.00 0.8381 0.7896 / 0.8421 / 0.8604 / 0.8604 0.819 0.7718 / 0.8235 / 0.8412 / 0.8412
truck 282 152 0.50/1.00/2.00/4.00 0.7756 0.5480 / 0.8328 / 0.8509 / 0.8707 0.760 0.5375 / 0.8190 / 0.8329 / 0.8520
bus 23 19 0.50/1.00/2.00/4.00 0.9979 0.9979 / 0.9979 / 0.9979 / 0.9979 0.989 0.9889 / 0.9889 / 0.9889 / 0.9889
bicycle 2 19 0.50/1.00/2.00/4.00 0.0000 0.0000 / 0.0000 / 0.0000 / 0.0000 0.000 0.0000 / 0.0000 / 0.0000 / 0.0000
pedestrian 324 123 0.50/1.00/2.00/4.00 0.3215 0.3065 / 0.3065 / 0.3201 / 0.3529 0.246 0.2339 / 0.2339 / 0.2452 / 0.2698

mAP: 0.6097, mAPH: 0.5846 (Plane Distance)

Label: car

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 586 337 0.8597 0.8407 0.8928 0.9688 0.8279 0.323492
4.00 586 337 0.8604 0.8412 0.8928 0.9688 0.8279 0.323492

Label: truck

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 282 152 0.8509 0.8329 0.8819 0.9338 0.8355 0.229696
4.00 282 152 0.8568 0.8388 0.8819 0.9338 0.8355 0.229696

Label: bus

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514
4.00 23 19 0.9979 0.9889 1.0000 1.0000 1.0000 0.849514

Label: bicycle

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 2 19 0.0000 0.0000 nan nan nan nan
4.00 2 19 0.0000 0.0000 nan nan nan nan

Label: pedestrian

Threshold Predict_num Groundtruth_num AP APH max_f1 optimal_recall optimal_precision optimal_conf
2.00 324 123 0.3201 0.2452 0.5193 0.4568 0.6016 0.225948
4.00 324 123 0.3529 0.2698 0.5488 0.4390 0.7317 0.181541

Summary:

Label Predict_num GT_nums Thresholds mean AP APs Mean APH APHs
car 586 337 2.00/4.00 0.8601 0.8597 / 0.8604 0.841 0.8407 / 0.8412
truck 282 152 2.00/4.00 0.8539 0.8509 / 0.8568 0.836 0.8329 / 0.8388
bus 23 19 2.00/4.00 0.9979 0.9979 / 0.9979 0.989 0.9889 / 0.9889
bicycle 2 19 2.00/4.00 0.0000 0.0000 / 0.0000 0.000 0.0000 / 0.0000
pedestrian 324 123 2.00/4.00 0.3365 0.3201 / 0.3529 0.258 0.2452 / 0.2698

@vividf vividf requested a review from Copilot December 25, 2025 08:42
@vividf vividf self-assigned this Dec 25, 2025
Copilot AI reviewed Dec 25, 2025
View reviewed changes
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR integrates CenterPoint into a unified deployment pipeline by creating a new project bundle under deployment/projects/centerpoint/ and removing legacy deployment scripts. The integration enables PyTorch, ONNX, and TensorRT export/evaluation workflows through a componentized multi-file ONNX approach.

Key Changes:

  • Centralizes CenterPoint deployment under a unified CLI (deployment.cli.main centerpoint)
  • Implements multi-file ONNX export (voxel encoder + backbone/head components) with per-component TensorRT engine conversion
  • Removes legacy deployment entrypoints (projects/CenterPoint/scripts/deploy.py, projects/CenterPoint/runners/deployment_runner.py, projects/CenterPoint/models/detectors/centerpoint_onnx.py)

Reviewed changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 9 comments.

Show a summary per file
File Description
projects/CenterPoint/scripts/deploy.py Removed legacy deployment script
projects/CenterPoint/runners/deployment_runner.py Removed legacy runner implementation
projects/CenterPoint/models/detectors/centerpoint_onnx.py Removed legacy ONNX detector variant
projects/CenterPoint/models/__init__.py Removed ONNX model imports/exports
projects/CenterPoint/README.md Updated with unified deployment CLI usage
projects/CenterPoint/Dockerfile Added deployment dependencies (onnxruntime-gpu, tensorrt-cu12)
deployment/projects/centerpoint/runner.py New unified deployment runner
deployment/projects/centerpoint/pipelines/*.py Backend-specific inference pipelines (PyTorch/ONNX/TensorRT)
deployment/projects/centerpoint/export/*.py ONNX/TensorRT export pipelines with component extraction
deployment/projects/centerpoint/onnx_models/*.py Relocated ONNX model variants with import path fixes
deployment/projects/centerpoint/*.py Core deployment components (entrypoint, evaluator, data_loader, model_loader)
deployment/projects/centerpoint/config/deploy_config.py Centralized deployment configuration
deployment/projects/centerpoint/cli.py CLI flag registration (--rot-y-axis-reference)

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@vividf vividf requested review from KSeangTan and yamsam January 9, 2026 04:00
@vividf vividf marked this pull request as ready for review January 9, 2026 04:00
@vividf vividf force-pushed the feat/new_deployment_and_evaluation_pipeline branch from 68d1700 to 7ac5b8b Compare January 29, 2026 08:08
@vividf vividf force-pushed the feat/centerpoint_integration branch from 0c61463 to a9cd830 Compare January 29, 2026 08:10
vividf and others added 15 commits January 29, 2026 18:39
@vividf
feat: integrate centerpoint to new deployment framework
3f9e293 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
Update projects/CenterPoint/README.md
11780c9 
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
@vividf
Update deployment/projects/centerpoint/pipelines/tensorrt.py
8e47c76 
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
@vividf
Update deployment/projects/centerpoint/onnx_models/pillar_encoder_onn…
bd6e54e 
…x.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
@vividf
Update deployment/projects/centerpoint/onnx_models/centerpoint_head_o…
8cdedff 
…nnx.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
@vividf
chore: removed comment
d4d5553 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: clean up centerpoint export pipeline
52f433d 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: rename backend config to trt config
4a26d9e 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
refactor: refactor deploy config
1ea169c 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: clean code
b52dc4f 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: integrate centerpoint with new component structure
ec9e97f 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: fix kwargs
6dfdc59 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: update metrics output
359bcbb 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: clean code
f7a90c1 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
chore: clean up kwargs
97d977b 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf vividf force-pushed the feat/centerpoint_integration branch from a9cd830 to 97d977b Compare January 29, 2026 09:40
KSeangTan
KSeangTan reviewed Jan 29, 2026
View reviewed changes
deployment/core/config/base_config.py
model_inputs: Tuple[TensorRTModelInputConfig, ...] = tuple(
TensorRTModelInputConfig.from_dict(item) for item in model_inputs_raw
)
def from_dict(cls, config_dict: Mapping[str, Any]) -> "TensorRTConfig":
Copy link
Collaborator

@KSeangTan KSeangTan Jan 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

use from __future__ import annotations for self-reference without ""

KSeangTan
KSeangTan reviewed Jan 29, 2026
View reviewed changes
deployment/core/config/base_config.py
max_workspace_size=config_dict.get("max_workspace_size", DEFAULT_WORKSPACE_SIZE),
)

def get_precision_policy(self) -> str:
Copy link
Collaborator

@KSeangTan KSeangTan Jan 29, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Adding @property is cleaner, and remove get_

Copy link
Collaborator Author

@vividf vividf Feb 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, fixed in another PR in b75687c

KSeangTan
KSeangTan reviewed Jan 29, 2026
View reviewed changes
KSeangTan
KSeangTan reviewed Jan 29, 2026
View reviewed changes
KSeangTan
KSeangTan reviewed Jan 29, 2026
View reviewed changes
deployment/core/config/base_config.py
self.export_config = ExportConfig.from_dict(deploy_cfg.get("export", {}))
self.runtime_config = RuntimeConfig.from_dict(deploy_cfg.get("runtime_io", {}))
self.backend_config = BackendConfig.from_dict(deploy_cfg.get("backend_config", {}))
self.tensorrt_config = TensorRTConfig.from_dict(deploy_cfg.get("tensorrt_config", {}) or {})
Copy link
Collaborator

@KSeangTan KSeangTan Jan 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do TensorRTConfig.from_dict return {}? Is so, we dont need or {}

Copy link
Collaborator Author

@vividf vividf Feb 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, fixed in 152ff54

@vividf
chore: temp fix for t4metric2
a9ecfdd 
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
@vividf
Copy link
Collaborator Author

vividf commented Feb 2, 2026 •
edited
Loading

Close this PR and seperate them to two new PRs. One for refactor and one for integrate centerpoint.
@KSeangTan
Please review them inorder, thanks

  1. feat(deployment): refactor config and clean code #180
  2. feat(deployment): centerpoint deployment integration #181

@vividf vividf closed this Feb 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@KSeangTan KSeangTan KSeangTan left review comments

Copilot code review Copilot Copilot left review comments

@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.

2 participants

@vividf @KSeangTan