alishobeiri / monocular-video-odometery Goto Github PK

I was wondering if anyone else has faced the same error and would be kind to share the solution.
I am running the code on KITTI dataset "07" and "00"
python v3.11.4
numpy v1.25.2
opencv-python v4.8.0.76

Monocular-Video-Odometery\monovideoodometery.py", line 110, in visual_odometery

, self.R, self.t, _ = cv2.recoverPose(E, self.good_old, self.good_new, self.R, self.t, self.focal, self.pp, None)
cv2.error: OpenCV(4.8.0) D:\a\opencv-python\opencv-python\opencv\modules\calib3d\src\undistort.dispatch.cpp:592: error: (-215:Assertion failed) npoints >= 0 && src.isContinuous() && (depth == CV_32F || depth == CV_64F) in function 'cv::undistortPoints'

As ypu developed this with the help of Avi Singh post Monocular Visual Odometry which is in C++ (https://github.com/avisingh599/mono-vo) as in your code of python you are getting the green and red path both as VO estimation and ground truth both , but when compare with Avi sing post of C++ he is only getting the ground truth path not the VO path .

Can you tell me what is the difference you made that you are getting both the paths .

Thanks

Hi, do you think the code on Line 118 of monovideoodometery.py should be as follows instead of the current one?

self.R = self.R.dot(R)

In this way, the MSE Error is also lower than the current result.

hi , i am trying to estimate x,y,z ,yaw ,pitch , roll from a drone video , i am using a single camera . so i want to estimate the location of the drone ie the elevation etc . can i use this repo or can you suggest the best path moving forward ? thanks

also what does Obtain scale information from an external source" mean ?

Alishobeiri, thanks for sharing your code.

How is the performance difference with the C++ implementation? Have you measured it? Any experience on this is appreciated.

Best, Volkan