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

how to train multi-view reconstruction? #57

Closed tiexuedanxin closed 3 years ago

tiexuedanxin commented 3 years ago

hello, If i want to train the network by myself to do multi-view reconstruction, like the angle, which .yaml file should i choose? Or should I use ours_depth_mvs_pretrained.yaml, and modify the test model file use the model i trained when do test.thank you very much.

m-niemeyer commented 3 years ago

Hi @tiexuedanxin , you can find the config files for the multi-view reconstruction experiments in the configs/multi_view_reconstruction subfolder. If you want to use a pretrained model, you can use the _pretrained config. Have a look at the instructions in the readme for more information. Thanks!

tiexuedanxin commented 3 years ago

Hi @tiexuedanxin , you can find the config files for the multi-view reconstruction experiments in the configs/multi_view_reconstruction subfolder. If you want to use a pretrained model, you can use the _pretrained config. Have a look at the instructions in the readme for more information. Thanks!

hello, I know where there is the config files for the multi-view reconstruction experiments in the configs/multi_view_reconstruction subfolder. and when want to use a pretrained model, can use the _pretrained config. I want to know which config file is your used to train the multi view reconstruction such as the angle、birds, because you say the pretrained config is can't used to train, I don't know if I train a new model of angle、birds use you data which config file to use and how to use the new trained model to generate mesh. Because I want to do multi view reconstruction, so i want to know which config file to use when train and generate mesh. thank you.