About the generalization capability of the proposed De-Bias Net

[LarryDong]

Hi, thanks for sharing the codes.

After reading your paper, I have some questions:

• In your paper, you trained the network on Blackbird and VID dataset, and tested on "unseen simple trajectoris". Are these test trajectoris in Blackbird and VID dataset?
• Do we need to first train the network on our own platform/sensor setup, if we've changed the platform or IMU mounting?

Thanks.

[JIANG-CX]

1. As stated in the paper, we did not train the network on the Blackbird or VID datasets; instead, we trained and tested it entirely on our own captured dataset, which is located in the dataset file.
2. Yes, the network should be trained on each specific platform/sensor setup.