Do you have a code with insertion Stanford Bunny into the scene?
Kakoedlinnoeslovo opened this issue · comments
I mean the code of implementation of Figure 1 of https://arxiv.org/pdf/1905.08598.pdf.
Thank you for reply!
Hello, thank you for your interest in the paper. Yes I have code somewhere on my computer however it is rather hard-coded at the moment, but here is the general idea if you want to generate the same figures (actually some explainations are in supplementary materials).
- Choose your image I1, and compute the estimated depth D1 with SharpNet
- Render a depth map D_bunny and an RGB image I_bunny of the Stanford Bunny using this ply model and a renderer of your choice (i used Blender), using an object pose of your choice.
- Compute the new RGB image I2 that is a copy of I1 but with
I2[D_bunny<D1] = I_bunny[D_bunny<D1]
This should put pixels of the rendered bunny image instead of original the image when the scene is occluded by the bunny, and vice/versa. - Look at your image and see if the bunny is at the distance you hoped it would be in the image, if not, repeat from 2.
I am sorry if you were looking more for a true augmented reality method, I may do that later for sequence of registered images but not in a little while.
Hello, Michael, thank you for your reply! Do you use contours on step 3? What do you mean by designation D1?
Hey, sorry for the late reply. No I do not use contours in step 3, only depth prediction. As for D1, it denotes the predicted depth map with SharpNet.