search

lyft / nuscenes-devkit

Devkit for the public 2019 Lyft Level 5 AV Dataset (fork of https://github.com/nutonomy/nuscenes-devkit)
Other
365 stars 103 forks source link

Converting Map to Vector Forms #94

Open ziqipang opened 2 years ago

ziqipang commented 2 years ago

Hi, Thank you for your wonderful dataset!

In short, I am wondering how we may convert the masked map into the form of vector maps? The vector maps here mean the HD-Maps in Waymo and Argoverse, which are used in SOTA methods like TNT[1].

The background of my question is that: I am trying to explore some motion prediction algorithms on the basis of Lyft perception dataset. (As the perception dataset has exhaustive bounding box annotations, which the motion prediction dataset does not have). Converting the HD-Map into the vectorized forms is necessary to using the SOTA algorithms such as TNT[1], VectorNet[2].

Thank you for your help!

[1] TNT: Target-driveN Trajectory Prediction [2] VectorNet: Encoding HD Maps and Agent Dynamics from Vectorized Representation

Alexanderisgod commented 1 year ago

Hi, have you down the work mentioned above, now i want to get the lane vector corresponding to the image view.

ziqipang commented 1 year ago

Ahhh sorry. I moved to other datasets for my research and have not tried converting a Vector format ever since. However, you may look at two papers called HDMapNet and VectorMapNet, I think they have some post-processing to convert a vectorized hd map.

Alexanderisgod commented 1 year ago

thank u for the reply.

---Original--- From: @.> Date: Sat, Sep 17, 2022 22:13 PM To: @.>; Cc: @.**@.>; Subject: Re: [lyft/nuscenes-devkit] Converting Map to Vector Forms (#94)

ziqipang commented 1 year ago

FYI I just put two more references to the reply above with editing; not sure if they are helpful. Good luck!