search

nkdnnlr / Event-Based-Camera-Simultaneous-Mosaicing-and-Tracking

12 stars 7 forks source link

Particle converges on point on trajectory, why? #7

Closed huetufemchopf closed 5 years ago

huetufemchopf commented 5 years ago
Bildschirmfoto 2019-05-03 um 17 41 06 Bildschirmfoto 2019-05-03 um 17 40 23
nkdnnlr commented 5 years ago

We estimated the three sigmas for the update step according to the rotation seen from the ground truth poses. For a batch size of 300, the resulting sigmas are sigma_x = 3.366398763318423e-05 sigma_y = 3.366410184326084e-05 sigma_z = 0.0005285784750737629

Because there seems to be not much movement if those are used, we use different multiplier on the sigmas. Below a zoomed in image of the output of 1000001 events, with multiplier 200. See log file below for more details. output_f200

There is not really movement happening either, also convergence on the yellow point. Trying different sigmas, but probably something else goes wrong.

quaternions_04052019T063735.log