Martin Brossard

The simple way is to run the filter after commenting the update step. En 20 oct. 2020 10:21, en 10:21, "宁缺毋滥" escribió: >Hello ， I want to know how to...

Hi, I did not made experiment with multiple IMUs, but I think you have two possibilities: - Keep the same approach and architecture, and add different IMU as inputs. The...

Can you please describe how to try to compute the Jacobian ? En 4 mar. 2021 4:40, en 4:40, chenwinki escribió: >Hi, thank you for your really greatful work. I...

Hi, this value is used to avoid to have big neural network outputs during the first iterations and thus to train a little bit faster. A pretty good value is...

I think if sufficient if you remove all the `.cuda()` in the code.

I have not tried to validate on the StructVIO or ADVIO datasets.

Hello Sharish, I recognize that invariant Kalman theory is not straightforward. My best suggestion for reading are: A. Barrau and S. Bonnabel, “[Invariant Kalman Filtering](https://www.annualreviews.org/doi/abs/10.1146/annurev-control-060117-105010),” Annual Review of Control, Robotics,...

Hi, I can suggest  you to start as simple as possible. If it is only a rosbag, you can manually correct for imu biaises and imu misalignment. You thus have...

Hi Sagi, Could you be please describe precisely where you meet difficulties ? Is it for understanding the filter, implementing it in e.g. ros, ... ? If the problem concerns...

You give an input of dimension 9 but the network expects an input of dimension 6. It is sufficient to modify the input dimension (the 6) [here](https://github.com/mbrossar/ai-imu-dr/blob/ddf2a52f758a10f5c20cc1649f747918bade766a/src/utils_torch_filter.py#L47).