how to estimate the geo-registered information for the custom dataset?

[amughrabi]

Thanks for the useful work!

I have a question about creating a custom LLFF-like dataset. Based on my understanding, we need to run colmap on the images that we have. Then, we need to use the VIO algorithm to predict the geo-registered information and group them in a text file (filename, X, Y, Z).

Can you please highlight how I can obtain the geo-registered information using VIO?

[Dominic101]

I can provide some guidance for this, but I have a couple questions to clarify that you are looking for:

When you say LLFF-like dataset, do you mean single shot localization tests like we did when experimenting on the LLFF dataset or do you intend to run a sequence of images where you want some odometry source to be used for the predict step (like we did for the robot experiment). If you want the first option, then this is easier because all we need to do is input ground truth scale to get metric scaled poses from colmap. If you want the second option then we need to make sure the VIO frame and get's aligned to the colmap frame which takes a couple extra steps.

[amughrabi]

Thanks a million for replying to my issue. The data was collected with a mobile phone, Huawei Mate 20 Pro, using a custom application implemented in Android Studio to record videos and sequences of inertial data from its internal IMU. Each video consists of a sequence of video frames obtained with a variable sampling rate of around 2 Hz and IMU data recorded at a fixed sampling rate of 100 Hz. This is what the data looks like:
https://ubarcelona-my.sharepoint.com/:u:/g/personal/ahmad_almughrabi_ub_edu/ESvdEvfVWn1OttGNE3F2bckBUKYaxdpU8YeseoyLlqFS3A?e=JrZeIc

I also tried to get a small scene from them to make tracking easier. The images I have are listed in cam0, while I have a set of IMU data (Accelerometer and Gyroscope) captured from Andriod in imu0. I also created a bag file using vio-common https://ubarcelona-my.sharepoint.com/:u:/g/personal/ahmad_almughrabi_ub_edu/EU2sYpJBKBhPmnWXkPkcexoBlIvPjeGHoP5zDfNlOYJUgw?e=S5qP8c