Merge Nerfstudio-gsplat v.1.5.3 into current codebase#3
Merged
haodongw101 merged 521 commits intomainfrom Jan 21, 2026
Merged
Conversation
* small improvements to nd speed
* Remove projection matrix Which was redundant with the pinhole intrinsics parameters fx, fy, cx, cy. Leave the projmat parameter to ProjectGaussians for backwards compatibility. * Remove projmat also from _torch_impl to fix tests
Allows camera pose optimization in Nerfstudio
* remove 'projmat' arg
* implement and profile new sh * rewrite fast sh * clean up * fix naming of tmp variable * reformat * clean and add comments * minor * profile sh * format * fix typo * auto diff check for fast sh * minor --------- Co-authored-by: Jianbo Ye <jianboye@amazon.com> Co-authored-by: Ruilong Li <397653553@qq.com>
* absgrad * add version bump --------- Co-authored-by: Ruilong Li <397653553@qq.com> Co-authored-by: Jianbo Ye <jianboye@amazon.com> Co-authored-by: Justin Kerr <justin.g.kerr@gmail.com>
* bg grad * delete lines for debug * . * only compute bg grad when needed * reformat
Co-authored-by: Ruilong Li <397653553@qq.com>
* enable doc versioning * minor fix --------- Co-authored-by: Ruilong Li <397653553@qq.com>
Co-authored-by: Ruilong Li <397653553@qq.com>
* move github button to left * fix brand --------- Co-authored-by: Ruilong Li <397653553@qq.com>
* move github button to left * fix brand * docs update --------- Co-authored-by: Ruilong Li <397653553@qq.com>
Co-authored-by: Ruilong Li <397653553@qq.com>
…dio-project#182) * Improve numerical stability of _torch_impl.project_cov3d_ewa * Linter fix: reformat _torch_impl.py
Link to evaluation page was broken, replaced with latest results
* fetch py3.7 * black check done
* init commit of cuda v2 upgrade * example script complete functionalities * support undistort * render_traj * render_traj with depth * add comments * minor readme * minor * minor cleanup * abs grad supported and benchmarked * more benchmarks on AbsGrad * clean up sparse grad * benchmark on U4 * fix setup.py * minor bug fix * bug fix * fix setup * fix import * sh fix * code structure cleanup * restructure examples * bug fix * bug fix * minor * fix glm import bug * fix glm import * fix import * fix relative path in setup.py * no abs path in setup * use abs path to find glm * not abs path * submodule * support antialiasing * fix for compensation init * update markdown * merge packed rast with unpacked * packed fulled merged in rasterize_to_pixels * merge packing for isect_tiles * sparse grad cleans up * sparse grad and supports unnorm quat * update trainer for sparse grad * doc for rasterization * example code in rasterization * docs update * migration instructions * docs * docs * docs * benchmark on mipnerf360 * benchmark done * fix timing in benchmark table * update benchmark * evaluation doc cleanup * cleanup doc * cleanup image fitting example * example doc for colmap * minor * a bit cleanup * simple viewer example * index doc update * bug fix for render_traj * switch to tyro * format * fix trainer * train script minor cleanup * minor cleanup trainer script * minor cleanup * minor cleanup * profiling update * profile script cleanup * profile update * profiling code update * full profiling * profiling doc * profile doc update * update doc * profile doc * update doc * profiling doc * cleanup doc * doc * minor cleanup * properly expose _functions * minor * minor * sh supports grad w.r.t dirs * cleans up #pragma unroll * doc cleanup for _torch_impl * doc cleanup for _wrapper * minor cleanup * doc for rasterization * minor doc cleanup * default to colors = torch.clamp_min(colors + 0.5, 0.0) * bump version * acknowledge nerfview * fix video auto scale * add contributors * update url * update migration doc * minor doc update * typos * set version to 0.1.12 * relax tests a little bit * set version to 1.0.0 * rename latest to main * cleanup scripts * fix berkeley typo * minor cleanup * black format * JIT only if necessary * doc update * support multi camera for colmap * support multi-camera * fix width & height for multi camera colmap * add exploration doc * update readme * cleanup readme v2 * test is only enabled with CUDA devices * minor * minor fix * bug fix for sh to rgb +0.5 * bug fix for K in eval * bug fix for multi camera * fix imsize for multi camera * minor modification on dataset class * colmap parser supports recursive rel path * format * minor * minor * bug fix for multi camera * minor * zero init compensations in packed projection * mark Optional inputs for projection * black format * nits * cleanup profiling code comments * _QuatScaleToCovarpreci -> _QuatScaleToCovarPreci * create_splats_with_optimizers return type fix * fix typo * colmap dataset minor * colmap parser image_id * --random_bkgd argument * support batch size training; cleanup scheduler * support --patch_size * format * enable pose optim * black format * minor cleanup * add cam opt and app opt modules * support appearance and pose opt in script * minor fix * add --global_scale * format * rename cindices and gindices * gauss_ids -> flatten_ids in isect_offset_encode * rename projection -> fully_fused_projection * rename _Projection-> _FullyFusedProjection * rename rasterize_to_indices_iter -> rasterize_to_indices_in_range * cleaning up warnings: int -> uint32_t * minor * renaming the rest of cindices/rindices * camera_ids and gaussian_ids: int32 -> int64 * rename: compute_means2d_absgrad -> absgrad * support tensorboard in simple trainer * format * minor * move ninja to necessary requirement * minor * profiling mem history * minor cleanup for warning * use DoubleBuffer for radixsort. mem efficient * memory_history includes backward * support depth loss * black format * cu file formats * minor dataset fix * Update nerfview scripts (nerfstudio-project#202) --------- Co-authored-by: Ruilong Li <397653553@qq.com> Co-authored-by: JunchenLiu <liujunchen0214@gmail.com> Co-authored-by: Brent Yi <yibrenth@gmail.com> Co-authored-by: Hang <hangg.sv7@gmail.com>
Fix log2 -> log2f to force it use cuda compatible function
* Update simple_viewer.py * Update gsplat_viewer.py
* Enforce eval3d with UT is set * Update simple_trainer.py * format
* Update requirements.txt adding fused_bilagrid * Update simple_trainer.py for accomdate fused_bilagrid * Fix bilateral grid when Batch Size > 1 * Fix wrong indentation * Fix wrong logic * Recover unimported lib_bilagrid * Refactor simple_trainer.py * Add logic for activating bilateral_grid when fused_bilagrid is activated * Apply black formatting to fix CI * format --------- Co-authored-by: liruilong940607 <ruilongli94@gmail.com>
* Update simple_viewer.py * Update gsplat_viewer.py * Create gsplat_viewer_2dgs.py * Update simple_trainer_2dgs.py * Create simple_viewer_2dgs.py * Update simple_viewer_2dgs.py * should use render_mode="RGB+ED"
* fix viewer * thin_prism_coeffs with shape 4 * fix comment * fix remaining comment
…fstudio-project#743) the outputs of torch.sigmoid and torch.exp are always positive, therefore we need not call torch.abs on their results.
* ftheta model ported in; projection_ut_3dgs_fused support * ftheta fully support. untested * 3dgut readme * fixes such that test_ftheta passes * format * minor fix * ftheta fix * format * cleanup
* Format setup.py * Update setup.py
Co-authored-by: Manuel Dahnert <ga67lif@tum.de>
…erfstudio-project#831) * Optimize packed viewdir pass by reducing many-to-one indexing operations Co-authored-by: AbhinavGrover <abhinav.grover@applied.co> * Copy the indptr to cpu to avoid GPU sync in the loop * Resolved PR comments * Format * Format * typo --------- Co-authored-by: AbhinavGrover <abhinav.grover@applied.co>
…studio-project-main
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.