7DoF and 6DoF about tsformer-vo

[aofrancani]

For this one I use https://github.com/Huangying-Zhan/kitti-odom-eval, they already save the plot and other analyses for you...

[ztx991028]

Thank you for your reply, can you publish a fully trained model, because the results I get using the pre-trained model you gave don't seem to be very good. I would like to validate this with the fully trained model you gave

[aofrancani]

I have already published the fully trained ones. As I understand from your other comments and issues, you are training the model further and might be overfitting to the training data. Just load and evaluate the model without starting training and you will get the same results I reported.

[ztx991028]

Thank you for your reply, I downloaded the VO and arg file, put it in the location shown in the picture, and then changed the name of the checkpoint, is that so, thank you

[ztx991028]

But what you get is this, what do you think is wrong, thank you for your reply.

[aofrancani]

This is the result without the alignment. Check https://github.com/Huangying-Zhan/kitti-odom-eval to see how to run the optimization. In particular, sequence 01 is challenging due to its high-speed scenario.

[ztx991028]

I pasted the predicted 01 .txt into the Kikit evaluation tool and executed the command as shown, but the result seemed to be wrong. Thank you very much for your reply, thank you

[ztx991028]

Thank you very much for your reply, I already get it.

[ztx991028]

The data in your paper needs the command to use 7DoF in the evaluation toolbox, right, thanks