search

autonomousvision / differentiable_volumetric_rendering

This repository contains the code for the CVPR 2020 paper "Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervision"
http://www.cvlibs.net/publications/Niemeyer2020CVPR.pdf
MIT License
794 stars 90 forks source link

DTU data used in DVR #56

Closed rohanchabra closed 3 years ago

rohanchabra commented 3 years ago

I wanted to know the reason why you guys used projection_matrices (provided in .txt) and not the calibration file (.mat). The issue is that in my project I want to use the original camera and pose matrices in .mat file and the scale_matrix provided by DVR to normalize the scenes.

m-niemeyer commented 3 years ago

Hi @rohanchabra , thanks for your interest in the project. It's true that we use the projection matrices; you can check out the FAQ if you want to have a detailed look at how we use the camera instrinsics and extrinsics in our model. If you want to directly decompose the projection matrix from DTU into "[Rt]", you can use opencv's deompose function.

Good luck for your research!