Dennis Cychuang

HorizonAD version do static stereo matching in **traceStereo()** in the scope of **setCoarseTrackingRef（）**.

Actually, HorizonAD didn't applied Stereo-DSO fully according to original Stereo-DSO paper. In my opinion , they treat static and temporal stereo matching as the same weight(i.e lamda = 1), which...

The λ factor in the objective function actually goes into the information matrix(W) in G-N normal equation (J^TWJ x = -JWr), so you can find the λ factor in `double...

It has nothing to do with Lagrange Multiplier. Since DSO is a direct method, it can't normalized the norm of visual and imu residual to the non-dimensional terms（ Mahalanobis distance）as...

The author derived the right to left Jacobian, which is also mentioned in [issue28](https://github.com/RonaldSun/VI-Stereo-DSO/issues/28). However, I have no idea how he derived that.

@tanjunyao7 Your derivation really makes sense, thanks a lot!

@RonaldSun Same question here, I'm also wondering how you derived the right to left Jacobian(J_r_l) for SE3? Is there any reference we can look up to? Thanks!

@gariel-google Very appreciate your detailed instructions, and I'll test it and report the results! But I've two little questions about inferring depth from EuRoC MAV dataset. 1. After reading the...

The initial position of visual odometry depends on the time finishing initialization, so if you need global position information, then pure VO is not enough for your need.

@Metalzero2 Yes, you are right, so basically we can't determine the home position (0,0,0) in visual odometry. By the way, are you still working on DSO? What project are you...