Skip to content
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.

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

Waymo evaluation issue #29

Open
mincheoree opened this issue Jan 22, 2024 · 6 comments
Open

Waymo evaluation issue #29

mincheoree opened this issue Jan 22, 2024 · 6 comments

Comments

@mincheoree
Copy link

mincheoree commented Jan 22, 2024

Hello, thanks for great works!
During Waymo dataset evaluation, I encountered 0 mAP and GT objects.
image
I searched in PCDet issue, the author says that Waymo 1.0.0 is incompatible with evaluation, and they use Waymo 1.2.0 dataset as default, referring to this: open-mmlab/OpenPCDet#1102

However, in your AD-PT paper, you mentioned that you used Waymo 1.0.0 as your experiment setting. As your codebase is based on PCDet, may I ask if you also encountered this 0 AP error during waymo evaluation? Is Waymo 1.0.0 okay for evaluation?

@BOBrown
Copy link
Collaborator

BOBrown commented Jan 22, 2024

@mincheoree
hi, here is my conda envs of waymo:
# Name | Version
waymo-open-dataset-tf-2-0-0 | 1.3.1

please consider using

pip3 install waymo-open-dataset-tf-2-0-0

plus, I found in your log that the parsing ground truth is [0,7] [0]. No GT is loaded, and it is recommended that you check the loaded waymo gt_infos_train.pkl

@mincheoree
Copy link
Author

@BOBrown
Thank you for providing specific environment version! I will try it out.
For your other suggestion, I cannot find gt_infos pickle.
In your DB readme, the waymo dataset is processed as below, and I could not find gt_infos_train.pkl file.
image

@BOBrown
Copy link
Collaborator

BOBrown commented Jan 22, 2024

@mincheoree
waymo_processed_data_v0_5_0_infos_train.pkl and waymo_processed_data_v0_5_0_infos_val.pkl

@mincheoree
Copy link
Author

@BOBrown After moving to waymo dataset version 1.2, the issue is solved.

@mincheoree
Copy link
Author

mincheoree commented Feb 2, 2024

@BOBrown @sky-fly97 I got another question, I trained SECOND model (without IOU branch) with 3% waymo dataset from scratch (without pretrained model) with provided yaml, waymo 3% dataset config, but got quite different result with your report in README.
Your result is as following:

Data amount Overall Vehicle Pedestrian Cyclist
SECOND (From scratch) 3% 52.00 / 37.70 58.11 / 57.44 51.34 / 27.38 46.57 / 28.28

While I get

Data amount Overall Vehicle Pedestrian Cyclist
SECOND (From scratch) 3% 48.94 /41.96 55.74/ 54.96 48.25 / 32.78 42.84 / 38.14

Was your test sample interval set to 5?
This is quite different performance.

@sky-fly97
Copy link
Collaborator

@BOBrown @sky-fly97 I got another question, I trained SECOND model (without IOU branch) with 3% waymo dataset from scratch (without pretrained model) with provided yaml, waymo 3% dataset config, but got quite different result with your report in README. Your result is as following:

Data amount Overall Vehicle Pedestrian Cyclist
SECOND (From scratch) 3% 52.00 / 37.70 58.11 / 57.44 51.34 / 27.38 46.57 / 28.28
While I get

Data amount Overall Vehicle Pedestrian Cyclist
SECOND (From scratch) 3% 48.94 /41.96 55.74/ 54.96 48.25 / 32.78 42.84 / 38.14
Was your test sample interval set to 5? This is quite different performance.

Hello,did you use VoxelBackBone8x as 3d backbone while using second? Note that we use VoxelResBackBone8x as the 3D backbone. Please change it at here.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants