Turbulent lane curves when mapping 3D lanes onto pointclouds.

[Eralien]

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:

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

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.

[ilnehc]

@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]

@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]

@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]

@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]

@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]

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

[ilnehc]

@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]

@ilnehc Thank you for the update!

[RicardLee]

@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.

[spyflying]

@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]

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