search

cvg / pixloc

Back to the Feature: Learning Robust Camera Localization from Pixels to Pose (CVPR 2021)
Apache License 2.0
735 stars 92 forks source link

how to get the extrinsics between the cameras #7

Closed angiend closed 2 years ago

angiend commented 2 years ago

thanks for your great work. i have download the robotcar from the url, and run the code "run_RobotCar.py" and get the result like this,

left/1418235223450115.jpg 0.11704475375741745 0.03295656318232015 -0.6387011805495836 -0.759786280822208 -125.37236785888672 -23.192873001098633 -7.491024494171143 rear/1418235392061935.jpg 0.7276782815863018 -0.5911740329291926 -0.23252374914543117 -0.2587088853928167 -100.58859252929688 -41.86554718017578 207.1205291748047 right/1418236138953803.jpg 0.6205097424969388 -0.6552427198975653 -0.20077227393455488 -0.3812022186540521 24.56517219543457 7.398570537567139 -221.69622802734375

i would like to know how to get the extrinsics between the cameras?

sarlinpe commented 2 years ago

The format of the text file is:

image_name qw qx qy qz tx ty tz

The pose is expressed as quaternion qw qx qy qz and camera translation tx ty tz in the COLMAP coordinate system, i.e. from the world to the camera frame. We use the Computer Vision convention: the X/Y/Z axes point left/down/forward.

angiend commented 2 years ago

@Skydes thank you very much,it very helpfull. i also want to know how to get the extrinsics between cameras? example,the extrinsics between the left camera and the right camera ,from the result, i could get the single camera pose respect to the world ,is right?

sarlinpe commented 2 years ago

Sorry for the late reply. The extrinsics between left and right cameras are not publicly available for RobotCar, but you can estimate them using rotation averaging on the reference poses.

tsattler commented 2 years ago

The extrinsics for the cameras with respect to the car are actually available together with the RobotCar Seasons dataset. From this, you can compute the relative poses between the left and right camera.

angiend commented 2 years ago

thank you very much. @Skydes @tsattler