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

Turbulent lane curves when mapping 3D lanes onto pointclouds. #23

Closed
Eralien opened this issue Jul 15, 2022 · 11 comments
Closed

Turbulent lane curves when mapping 3D lanes onto pointclouds. #23

Eralien opened this issue Jul 15, 2022 · 11 comments

Comments

@Eralien
Copy link

Eralien commented Jul 15, 2022

Hi! Thank you for your great work. I am trying to map your 3D points onto the corresponding pointclouds of waymo dataset. I noticed a problem:

image

Certain lanes seem to have very turbulent curves, especially curbs as shown in purple.

image

I have checked the raw data and the z values do seem to be not smooth from point to point. I am not sure if any post-processing is required, or certain types of lane are not smooth by nature.

@Eralien Eralien changed the title Wrong extrinsics when mapping 3D lanes onto pointclouds. Turbulent lane curves when mapping 3D lanes onto pointclouds. Jul 15, 2022
@ilnehc
Copy link
Collaborator

ilnehc commented Jul 16, 2022

@Eralien Thank you for raising the issue. Could you provide the segment or json number that we can look into?

Btw, I notice that you have an issue related to the coordinate system. Have you already solved that?

@Eralien
Copy link
Author

Eralien commented Jul 16, 2022

@ilnehc Thank you for your quick response! The segment I'm using is segment-1005081002024129653_5313_150_5333_150_with_camera_labels, and the strange curves basically exists throughout the segment.

The other issue realted to the coordinate system was due to a stupid bug I wrote. It's been solved.

@nicoduchene
Copy link

@Eralien quick question. I imagine that the names of the json files are made from a timestamp.
Could you help me by letting me know which timestamp the json file names are built from?
Thanks!

@Eralien
Copy link
Author

Eralien commented Jul 25, 2022

@nicoduchene These timestamps were generated from the waymo dataset. The raw waymo dataset are composed with segment**.record files, with timestamp of each frame recorded. Here's a snippet I used:

dataset = tf.data.TFRecordDataset(str(sequence_file), compression_type='')
for cnt, data in enumerate(dataset):
            frame = dataset_pb2.Frame()
            frame.ParseFromString(bytearray(data.numpy()))
            timestamp = "{:<018}".format(frame.timestamp_micros) # openlane seems to use 18 digit

@Eralien
Copy link
Author

Eralien commented Jul 25, 2022

@nicoduchene the openlane dataset are organized in the same way as the decompressed waymo dataset, so I simply exported the pointcloud as pcd files in the same way, and name them with 18-digit timestamp.

@Eralien
Copy link
Author

Eralien commented Aug 4, 2022

@ilnehc Is there any follow up on this issue? Or is this an unavoidable flaw in your dataset?

@ilnehc
Copy link
Collaborator

ilnehc commented Aug 4, 2022

@Eralien We are working on it. Due to the large size of the dataset and limited time with other projects on hand, a huge effort is needed to make sure all similar cases are solved. Please be patient and we will publish a renewed version in about 2-3 weeks.

@Eralien
Copy link
Author

Eralien commented Aug 8, 2022

@ilnehc Thank you for the update!

@RicardLee
Copy link
Contributor

RicardLee commented Sep 1, 2022

@Eralien Hi, for these strange curves, we updated the dataset Lane3d_1000_V1.2. We added some smoothing algorithms for the XY of 3D lane to eliminate this strange curves. The new data can be obtained from baidu cloud Link.Shown in the following figures, red line for raw laneline, blue line for new laneline and we can find that a lot of noise data has disappeared.

@RicardLee RicardLee pinned this issue Sep 1, 2022
@spyflying
Copy link

@RicardLee Since the download of renewed annotation from google drive always fails, could you kindly provide the link of Baidu Cloud? Thanks a lot for the nice work.

@RicardLee
Copy link
Contributor

RicardLee commented Sep 7, 2022

@spyflying Hi, we have updated the baidu cloud link.

@faikit faikit mentioned this issue Mar 19, 2023
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

5 participants