Closed navidnadery closed 2 years ago
That depends on how you decide the foot points of the two objects. The method we used based on bounding boxes may not work well for high precision. I would suggest to choose the center of mass of the segmented mask. But the most accurate approach might be training an additional model to estimate the foot points of vehicles.
Thank you for your recommendations. I've tried centering of mass, but it does not help. However, training a model seems good.
Hi, Thank you for the tutorial. I have a problem in estimating real distance of 2 objects (an object in 2 consecutive frames). Height of the vehicle (from ground) can influence on our estimate, how can I handle it? In step 2, you use camera calibration then perspective, however they assume that all object are on ground and are flat.
I really appreciate your guidance