R and t looks wrong

[mnauf]

I am using D415 Realsense camera, with following intrinsic parameters:

[[952.828   0.    646.699]
 [  0.    952.828 342.637]
 [  0.      0.      1.   ]]

Therefore, my records in scannet_sample_pairs_with_gt.txt looks like: First Record

100.jpg 161.jpg 0 0 952.828 0. 646.699 0. 952.828 342.637 0. 0. 1. 952.828 0. 646.699 0. 952.828 342.637 0. 0. 1. 0.15102 0.52115 -0.84 1.95984 -0.41332 0.80519 0.42525 -1.02578 0.89798 0.28297 0.337 1.24882 0. 0. 0. 1.

Second Record

161.jpg 100.jpg 0 0 952.828 0. 646.699 0. 952.828 342.637 0. 0. 1. 952.828 0. 646.699 0. 952.828 342.637 0. 0. 1. 0.15102 0.52115 -0.84 1.95984 -0.41332 0.80519 0.42525 -1.02578 0.89798 0.28297 0.337 1.24882 0. 0. 0. 1.

(Please note that after K0 and K1, random numbers are written and they aren't the ground truth, because I am not interested in calculating error.)

I did a sanity check whether Inverse of [R12, t12;0 0 0 1] is equal to [R21, t21;0 0 0 1] and it turned out to be pretty close, therefore good job on that.

First Question

My first question is from the "first record". Here, is 100.jpg an anchor image? and the resultant R and t allow us to go from 100.jpg to 161.jpg or it's the opposite?

Second Question

My second question is, the R and t for the first record is,

R:  [[ 0.99932221 -0.03031383 -0.02088506]
 [ 0.03172734  0.99696792  0.07105174]
 [ 0.01866788 -0.07166621  0.99725396]]
t:  [-0.07450811  0.13622056 -0.98787271]

while for second,

R:  [[ 0.99921875  0.03247319  0.02252493]
 [-0.03083453  0.9970961  -0.06963189]
 [-0.02472069  0.06888294  0.99731842]]
t:  [ 0.07328248 -0.21668444  0.9734873 ]

is tz for the first record i.e -0.98, which is the second index of t, is it in meters? Because if that's true, I barely took a few steps forward from 100.jpg to 161.jpg

Reference Images

The reference images are: 100.jpg 161.jpg