search

weitong8591 / ars_magsac

The code for ICCV 2023 paper, Adaptive Reordering Sampler with Neurally Guided MAGSAC
BSD 3-Clause "New" or "Revised" License
20 stars 1 forks source link

demo.py #3

Open ABCDZLG opened 1 month ago

ABCDZLG commented 1 month ago

捕获 Hello teacher, in the process of reproducing your code, there is an error in the picture, I made changes in the code, added a line ：correspondences = correspondences.numpy() to make a conversion, but still will report an error, may I ask the teacher specifically how to solve it? Thank you!

ABCDZLG commented 1 month ago

I was looking for another article of yours: differentiable_ransac, and saw a question on github with the same problem, although you replied on GitHub, I still don't know how to change the code, I hope you can guide me.Thank you！

Link to your answer in differentiable_ransac：https://github.com/weitong8591/differentiable_ransac/issues/9

weitong8591 commented 1 month ago

hi, there are different args in the newest magsac++ repo and mine, pls change the input as magsac++ (you installed) requires. For example, you should have the first input in a shape of n4 (x1, y1, x2, y2) instead of n2.

ABCDZLG commented 1 month ago

Thank you for your answer, but I have other questions I would like to ask you about: 1:What is the connection between model fitting and E 、 F matrix estimation?Is E and F estimation an application of model fitting? How to understand it? 2 Can I understand it like this: suppose a curve is fitted given some points in a 2D plane, treat it as a whole as A, then rotate this plane angle (R), translate (T) and get B. Estimate the E and F matrices by making a match between the two wholes A and B? 3 If the input is a matching correspondence on the image (NG-RANSAC), is this correspondence one-to-many? Or many-to-many? The final weights obtained through the neural network output are the weights of the data points? Or is it the weight of the line that the data points are connected to? I didn't understand it during the reading of the paper, can you help me to answer it?