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

Can I use your model to train or infrence trajectory on KITTI dataset? #30

Open
sunnyHelen opened this issue Dec 9, 2019 · 5 comments

Comments

@sunnyHelen
Copy link

Do you ever try to train or test on KITTI dataset? Can I use your model to predict trajectory on other dataset except 7 Scenes and RobotCar?

@sunnyHelen sunnyHelen changed the title Can I use your model to train or refrence trajectory on KITTI dataset? Can I use your model to train or infrence trajectory on KITTI dataset? Dec 9, 2019
@samarth-robo
Copy link
Contributor

Hi @sunnyHelen , yes you can use MapNet for other datasets -- create your own dataloader by following the ones already provided. Here is some documentation.

@sunnyHelen
Copy link
Author

Nice, thanks a lot~

@sunnyHelen
Copy link
Author

Hi, I created a data loader KITTI dataset. But when I visualize, I find even the trajectory of ground-truth are totally wrong. Like this:
image
And here's the description of the ground-truth pose:
image
I just used this ground-truth for visualization. Is this wrong? Or do you know what may cause this result?

@sunnyHelen
Copy link
Author

The right trajectory should be like this:
image

@samarth-robo
Copy link
Contributor

samarth-robo commented Dec 25, 2019

Hi @sunnyHelen it is hard to debug this without looking at your dataloader code. It seems like you should be able to load the poses file directly into the ps variable in your dataloader. I suggest to first turn all normalization off by setting mean_t=[0, 0, 0], std_t=[1, 1, 1], align_R = np.eye(3), align_t = np.zeros(3), align_s = 1, and seeing if that produces a reasonable plot.

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

2 participants