search

tinghuiz / SfMLearner

An unsupervised learning framework for depth and ego-motion estimation from monocular videos
MIT License
1.96k stars 555 forks source link

Image Inverse Warping #54

Open madhubabuv opened 6 years ago

madhubabuv commented 6 years ago

Hello, Thanks for the code, I am not able to understand the pose estimation. Suppose if I consider t and t+1 images, the predicted pose result should be positive in (+ve) in z-axis but I am getting -ve value. The same is happening with t-1 to t frames. Correct me if I am wrong, If you are using the predicted pose from the target (t) to source image (t+1) to reconstruct the target image by inverse warping, the predicted pose should be always positive right? (since the car is moving forward). Why the code showing -ve values for Z ?

pleasegostraight commented 4 years ago

Hi, I meet the same problem, how do you solve it?