Yuanchen Guo
Yuanchen Guo
Do you guys have any plan on releasing the pretrained models? Excellent work by the way.
- [x] scannet dataset @pablovelaimmersed - [ ] nerf/neus training on scannet dataset - [ ] normal/depth priors
- [x] score debiasing from [Debiasing Scores and Prompts of 2D Diffusion for Robust Text-to-3D Generation](http://arxiv.org/abs/2303.15413) **implemented as `grad_clip` in guidance** - [ ] prompt debiasing from [Debiasing Scores and...
- remove redundant config saving in `launch.py` - change `lambda_normal_smooth` and `lambda_3d_normal_smooth` to `return_comp_normal` and `return_normal_perturb` in `nerf-volume-renderer`
- Finalize ProlificDreamer implementation: all three stages tested. - Easy-to-use geometry conversion by `system.geometry_convert_from`. This is used in the Magic3D and ProlificDreamer system and may inspire applications connecting multiple systems/algorithms....
512x512 ProlifcDreamer will cost >40GB VRAM at the beginning of training, as we sample 512 points per ray, which is far more than 64 coarse + 32 fine (92 in...
This PR is related to #175 . I made the following changes: - add `set_min_max_steps` to all guidance (as in #175) - in the `update_step` hook of all guidance, use...
This PR is from [this PR](https://github.com/threestudio-project/threestudio/pull/173). I've made the following changes: - The MLP network is removed from the material, i.e., the material simply explains what the features given by...