jiawei-mo / scale_optimization Goto Github PK

Hello.
Thanks for sharing the work! I am trying to recreate the scale optimization in Matlab, to integrate into our Visual Odometry system. Could you point out the Jacobian calculation in the code? I checked out ScaleOptimizer and ScaleAccumulator, but could not find where and how the Jacobians are calculated (nor the reprojection errors, though implementing that is quite trivial) I did go through the paper, and seem to be stuck at finding Jacobians.

From what I could gather, for a given 3D point, the Jacobian in the y direction is the image gradient in the y direction times the ∂Π / ∂s in the 'y' direction. Similarly the values are found for the 'x' direction, and finally, are they added together, i.e. is Js the sum of Jacobian components in the x and y direction? These Js values are found for every point and finally, and using a Gauss Newton step, the scale increment is found.

Could you please elaborate on finding the Jacobian? (preferably with a tiny example)

I'm not entirely sure if this issue belongs here, so please feel free to delete it if inappropriate.
Thanks!

Hello,I download the project and try it with MH_05 dataset, but after it run just a second the core dump comes.Is it the same as you?

hello，can you tell me how to set scale_accept_th,I see you set it to 10 or 20,why?Thank you!

Hello,Would you please give your results files on euroc or kitti datasets contain no loop-closure and loop-closure?Or maybe you don't have all files,but if you have some files,thats ok. I am really need them,thank you very much！

Thank you very much for revealing a good algorithm.

I want to run it using my own dataset.
The build is done and the program running ok.

But I don't know where to place the image files.
Can you give a simple example?

param name="cam0_topic" value="[CAM0 TOPIC]
param name="cam1_topic" value="[CAM1 TOPIC] <- Is it a image folder path?