v1.1.5

What's Changed

• Suppress wandb type error by @brentyi in #3397
• Bump viser to 0.2.7 by @ginazhouhuiwu in #3398
• change grid resolution type by @hoanhle in #3400
• Add eval data to image_rename_map by @jiaruiyu99 in #3396
• update documentation for instant-ngp by @hoanhle in #3406
• Update dockerfile and add docker build action by @jkulhanek in #3283
• RGBA renderer for splatfacto by @hardikdava in #3307
• bugfix: handle multiple camera types in a single batch by @decrispell in #3415
• Add option to start training paused by @aayushg55 in #3420
• Fix docker attestation fail on forked repo PRs by @jkulhanek in #3423
• use gsplat strategy interface to simplify splatfacto by @jb-ye in #3376
• Minor update to generate splat files in inference mode by @maxwildersmith in #3432
• Fix crop obb masking in pointcloud generation by @nepfaff in #3433
• Adding Random Seed for Frame Processing by @Anthony-Tafoya in #3416
• Fix splatfacto crash in eval when using viewer by @aayushg55 in #3430
• Zipnerf point cloud and tsdf mesh export support by @aayushg55 in #3429
• Fix #3417 by @jb-ye in #3441
• Change depth from float64 to float32 by @nepfaff in #3453
• Speed up PairPixelSampler by @nepfaff in #3452
• Rejection sample mask for patch pixel sampler by @nepfaff in #3457
• Make PixelSampler the depth-nerfacto default by @nepfaff in #3454
• Keep depth_image on CPU by @nepfaff in #3449
• Add max_frames + max_output_size arguments for Aria data processing by @brentyi in #3466
• Add option for refining the TSDF using the mesh's AABB by @nepfaff in #3459
• Add option to change .ply output filename for Gaussian exports by @brentyi in #3467
• tag exported ply files by @simonbethke in #3472
• update gsplat version to >=1.3.0 by @kerrj in #3478
• Properly reset train state after render lock in viewer by @kerrj in #3486
• Restrict pymeshlab version on Windows (3461) by @SharkWipf in #3474
• fix: saving ckpt error when ckpt folder has some folder inside by @Crescent-Saturn in #3516
• bump version to 1.1.5 by @jb-ye in #3521

New Contributors

• @jiaruiyu99 made their first contribution in #3396
• @aayushg55 made their first contribution in #3420
• @maxwildersmith made their first contribution in #3432
• @Anthony-Tafoya made their first contribution in #3416

Full Changelog: v1.1.4...v1.1.5

v1.1.4

Bilagrid + Splatfacto/Nerfacto, and more bug fixes!

What's Changed

• Compatibility for numpy 2.0 by @kerrj in #3265
• Fix different initial tension value between tension_slider and `cam… by @yzslab in #3271
• Fix parsing of non ascii image names in colmap's images.bin by @TamirCohen in #3256
• Bump viser by @ginazhouhuiwu in #3234
• Update pytorch in Dockerfile to 2.1.2 by @Zunhammer in #3268
• Fix dimension mismatch between downscaled cameras and images in colmap. by @jisung3d in #3232
• Bump viser to 0.2.3 by @brentyi in #3277
• Make awscli an optional dependency by @pculbertson-bdai in #3279
• Document updates for splatfacto-w by @KevinXu02 in #3281
• fix color shape when sh_degree=0 in splatfacto by @liruilong940607 in #3296
• Update the docstring of SplatfactoModel.get_outputs by @kstoneriv3 in #3289
• Update installation.md: Add missing Pixi documentation by @SharkWipf in #3302
• Added _load_3D_points to scannet_dataparser.py by @junyuan-fang in #3305
• Loose assert condition for datamanger in get_average_eval_image_metrics() by @KevinXu02 in #3310
• Splatw doc updates by @KevinXu02 in #3311
• Pass current camera idx in metadata to get output from viser by @KevinXu02 in #3293
• fix ruff check by @jb-ye in #3328
• Swap out opencv-python for opencv-python-headless by @SharkWipf in #3336
• make np.array compatible with numpy 2.0 by @Guangyun-Xu in #3339
• Allows rendering the appearance of specific training camera indices by @KevinXu02 in #3346
• Lazy import for open3d in Blender DataParser by @ethanweber in #3347
• Bilagird for splatfacto by @KevinXu02 in #3316
• Bump + run ruff by @brentyi in #3369
• installation.md: Remove manual gsplat installation for Windows by @SharkWipf in #3333
• bug fix for splatfacto by @jb-ye in #3367
• Addressed TODO in full_images_datamanager.py to return image resolution as train rays by @ajean757 in #3312
• fix bugs in generfacto, train_dataparser_outputs.save_dataparser_transform and viewer camera on cpu by @pierremerriaux-leddartech in #3378
• Fix memory leak in full_images_datamanager.py by @brentyi in #3375
• Fix #3355 by @jb-ye in #3381
• Add an example on how to customize dataparser with ns-train by @yufengzh in #3385
• get rid of gsplat legacy dep by @liruilong940607 in #3386
• Updating camera intrinsics (principal points) when the undistorted image is cropped by @AntonioMacaronio in #3382
• add assertion to make sure positions and positions_flat are not empty by @hoanhle in #3387
• Update exporter.py to export sh_degree 0 case #3371 by @jb-ye in #3374
• bump version to v1.1.4 by @jb-ye in #3394

New Contributors

• @TamirCohen made their first contribution in #3256
• @jisung3d made their first contribution in #3232
• @pculbertson-bdai made their first contribution in #3279
• @kstoneriv3 made their first contribution in #3289
• @junyuan-fang made their first contribution in #3305
• @yufengzh made their first contribution in #3385

Full Changelog: v1.1.3...v1.1.4

v1.1.3

Bug fixes!

What's Changed

• Making install_checks use colmap_cmd by @SimonF24 in #3204
• Fix bug in SplatfactoModel when sh_degree is zero the color is not with sigmoid by @liruilong940607 in #3207
• Update installation guide for Windows (tcnn+gsplat) by @KevinXu02 in #3212
• Update installation.md by @KevinXu02 in #3213
• Adds back screen size too bigs culling as an option by @KevinXu02 in #3211
• Small fix when evaluating splatfacto model, running ns-eval by @ajean757 in #3219
• Add descriptions for splatfacto-big and nerfacto-huge by @KevinXu02 in #3233

New Contributors

• @SimonF24 made their first contribution in #3204
• @ajean757 made their first contribution in #3219

Full Changelog: v1.1.2...v1.1.3

v1.1.2

Fixes splatfacto for Windows.

What's Changed

• torch.compile -> torch_compile from misc by @liruilong940607 in #3202

Full Changelog: v1.1.1...v1.1.2

v1.1.1

What's Changed

• Fix legacy tensor construction by @gtangg12 in #3145
• Support OCI standard dockerfile by @SwirtaB in #2961
• Only import pycolmap in run_hloc by @KevinXu02 in #3146
• Add Pixi as an option for installation by @pablovela5620 in #3130
• remove redundant code in full_images_datamanager.py by @paolovic in #3155
• Add OpenNeRF method (ICLR'24 http://arxiv.org/abs/2404.03650) by @francisengelmann in #3167
• Fix "poster" dataset extraction by @martin-kokos in #3164
• Fixing resize_image and adding additional distortion check by @AntonioMacaronio in #3162
• NeRFtoGSandBack by @Siming-He in #3150
• Fix: when --use-sfm-depth is enabled and the model is given by --colm… by @Serenitysmk in #3172
• Export Splatfacto-optimized poses by @oseiskar in #3116
• Introduce FPSampler for train cameras by @jb-ye in #3177
• Generalize get average image metrics by @hoanhle in #3188
• Adding new captures to eyeful dataset by @isrish in #3186
• Ignore gsplat 1.0 for now by @pwais in #3197
• Update splatfacto with gsplat 1.0 by @liruilong940607 in #3200

New Contributors

• @gtangg12 made their first contribution in #3145
• @SwirtaB made their first contribution in #2961
• @paolovic made their first contribution in #3155
• @francisengelmann made their first contribution in #3167
• @martin-kokos made their first contribution in #3164
• @Siming-He made their first contribution in #3150
• @Serenitysmk made their first contribution in #3172
• @hoanhle made their first contribution in #3188
• @isrish made their first contribution in #3186

Full Changelog: v1.1.0...v1.1.1

v1.1.0

What's Changed

• Refactor ScenePointerCallbacks for viser compatibility by @chungmin99 in #3036
• Only load+undistort splatfacto images when they're used by @brentyi in #3043
• New Google Drive link for poster dataset by @ginazhouhuiwu in #3045
• Added SIGNeRF to Nerfstudio methods by @JDihlmann in #3053
• Make colmapDataParser compatible with 360_v2 dataset format by @Jing1Ling in #2860
• Fix config loading for Eyeful Tower dataset by @hardhat5 in #3054
• feat(ci): migrate to uv in CI by @SauravMaheshkar in #3038
• Fix custom elements nested under @Property by @kerrj in #3066
• fix exporter.py for FullImageDataManager and RandomCamerasDataManager by @dberga in #3065
• Camera pose optimization for Splatfacto by @oseiskar in #2891
• write eval combined images for diagnostics by @jb-ye in #3070
• Fixed Unexpectedly huge RAM consumption of parallel datamanager by @hyperzy in #3060
• Add pointcloud loading for blender parser by @gchhablani in #2902
• shift principle points by .5 in undistortion by @kerrj in #3071
• Fix call to detect ODM reconstruction file by @codeprimate in #3018
• updates to apply_pca_colormap by @kerrj in #3086
• Changes for trainer.py to support the Gradio webui by @KevinXu02 in #3046
• Allow alpha transparency training for blender datasets input by @jb-ye in #3088
• Hloc fix for image_ids (camera per image) by @dberga in #3081
• Adding lock for Viser Element callbacks by @Satvik1701 in #3067
• Adding cropping for gaussian splat exports by @THE-COB in #3105
• GaussianSplats export: allow uint8 in ply by @mathijshenquet in #3111
• Use AbsGrad to get better results with less gaussians by @jb-ye in #3113
• Fix bug in pixel sampler by @carlinds in #3103
• metashape_utils: work with camera groups by @f-dy in #2626
• Fix #3064 and some minor changes by @jb-ye in #3121
• Support splatfacto training for metashape exports by @jb-ye in #3122
• fix nerfstudio collate function by @hturki in #2965
• Add a pointcloud export from polycam to process-data by @kerrj in #3124
• BioNeRF by @Leandropassosjr in #3125
• Fixing typos and symbols in bionerf.md by @Leandropassosjr in #3126
• Add kwargs handling for patchpixelsampler patch size by @chungmin99 in #2622
• Make RealityCapture csv/ply exports work! by @jb-ye in #3131
• adding the link for the paper page and fixing image paths by @Leandropassosjr in #3132
• Fix edge case of depth colormap by @Feez2403 in #3139
• Clarify flag name in docs/nerfology/methods/splat.md by @EliasWatson in #3140
• bump version to 1.1.0 by @jb-ye in #3143

New Contributors

• @hardhat5 made their first contribution in #3054
• @dberga made their first contribution in #3065
• @codeprimate made their first contribution in #3018
• @Satvik1701 made their first contribution in #3067
• @carlinds made their first contribution in #3103
• @Leandropassosjr made their first contribution in #3125
• @Feez2403 made their first contribution in #3139
• @EliasWatson made their first contribution in #3140

Full Changelog: v1.0.3...v1.1.0

v1.0.3

What's Changed

• splatfacto-big by @AdamRashid96 in #2945
• Support new MPS filenames for Project Aria point clouds by @brentyi in #2949
• Show viewer resolution as WxH instead of HxW by @brentyi in #2948
• Fix Nerfacto + Aria by @brentyi in #2950
• Fix progress bar for full images datamanager by @brentyi in #2969
• fix bug that opacity compentation not being applied to depth raterization by @jinhwanlazy in #2966
• add splatfacto-big docs by @AdamRashid96 in #2955
• Adding Terminal Window to Jupyter Cells so all Colab users (free and pro) can train with Nerfstudio by @AntonioMacaronio in #2870
• Refactor splatfacto gaussian params with a ParameterDict by @kerrj in #2946
• doc: fix code typo in 'NeRF' method doc by @Crescent-Saturn in #2978
• Fix depth-nerfacto edge cases by @ethanweber in #2939
• Handle alpha images properly for instant-ngp and tensorf models by @trinh31201 in #2979
• Fix: Masks having wrong dimension after undistortion (causes crash in splatfacto) by @Zunhammer in #2987
• Camera models by @wooni-github in #2997
• Add 'within' function to SceneBox by @tobiasfshr in #2991
• enable coarse-to-fine training by @jb-ye in #2984
• Upgrade to latest nbformat 5.9.2 by @ringohoffman in #2964
• Make ply export property ordering backward compatible (fixes #3004) by @mathijshenquet in #3005
• Fix rotation_matrix for opposite float32 vectors by @jb-ye in #3011
• Fix bug when trying to create a camera path in viewer by @MildMystery in #3016
• Support rectangle selection in viewer by @chungmin99 in #3014
• Get rid of redundant projection matrix computations in Splatfacto (needs a gsplat update) by @oseiskar in #3019
• Exporting Splatfacto: correct errors when there are NaN/Inf values in some attributes by @MaximeVandegar in #3032
• Fix core tests with Pillow>=10.3.0 by @jb-ye in #3034
• Support COLMAP reconstructions with non-flat image dirs by @zacstewart in #3006
• bump version to 1.0.3 by @jb-ye in #3035

New Contributors

• @Crescent-Saturn made their first contribution in #2978
• @trinh31201 made their first contribution in #2979
• @wooni-github made their first contribution in #2997
• @mathijshenquet made their first contribution in #3005
• @MildMystery made their first contribution in #3016
• @MaximeVandegar made their first contribution in #3032
• @zacstewart made their first contribution in #3006

Full Changelog: v1.0.2...v1.0.3

v1.0.2

What's Changed

• fix: pixelated downscaled images by @ymtoo in #2866
• Add multi-processing support for cache_image. Fixed the issue of insufficient precision leading to a completely black image when converting from RGBA to RGB. by @KevinXu02 in #2867
• Enable always returning depth for Gaussian Splatting by @ethanweber in #2873
• Update viewer legacy docs by @brentyi in #2875
• Add default depth output when no gaussians visible by @chungmin99 in #2874
• Remove useless call to .with_suffix(""), which causes a bug if filename contains dots by @f-dy in #2882
• support generic TDataset in paralleldatamanager by @hturki in #2884
• [Fix] convert camera_idx to tensor by @tobiasfshr in #2890
• Fix benchmarking script camera optimizer argument by @gchhablani in #2899
• add zipnerf-pytorch as external method by @Jing1Ling in #2850
• Add transparency carving to splatfacto by @pa-la in #2889
• Fix an issue where gradient accumulation could not be passed as argument due to a type error. by @SharkWipf in #2910
• Implement new rasterization mode using opacity compensation factor by @jb-ye in #2888
• Switch default matching method for videos to 'sequential' in COLMAP by @kerrj in #2912
• Call rescale_output_resolution before early return by @pierotofy in #2897
• Make time slider effective in viewer (Fixes #2928) by @iSach in #2931
• Fixed ns_render dataset bug in splatfacto by @pa-la in #2917
• Update project, rasterize with block_width interface by @kerrj in #2937
• Full implementation of features for dynamic scenes into the Viser viewer. by @iSach in #2936
• fix crash when both mask and downscale_factor are set by @jinhwanlazy in #2920
• 1.0.2 by @kerrj in #2943

New Contributors

• @gchhablani made their first contribution in #2899
• @Jing1Ling made their first contribution in #2850
• @pa-la made their first contribution in #2889
• @jinhwanlazy made their first contribution in #2920

Full Changelog: v1.0.1...v1.0.2

v1.0.1

Bug fixes!

What's Changed

• important fix typo by @jb-ye in #2800
• Add documentation on existing datasets by @VasuAgrawal in #2806
• Update splatfacto.py by @Harr7y in #2804
• Fix typo in parallel_datamanager.py by @eltociear in #2812
• Make downloading more windows-friendly by @VasuAgrawal in #2810
• Have Mill19 dataset use official splits by default by @hturki in #2745
• Fix mask shape for Fisheye624 cameras by @brentyi in #2813
• Use packaging for colmap version by @chungmin99 in #2814
• Remove viewer build badge by @brentyi in #2819
• Add alpha rendering to splatfacto, fix depth bug by @kerrj in #2824
• Fix "mutable default" error in render.py by @pyetras in #2825
• fixing colmap installation to use apt-get since gdown on the URL unable to download colmap by @AntonioMacaronio in #2823
• Update viewer docs by @brentyi in #2826
• Fix the background color bug when exporting video with the "random" background option by @KevinXu02 in #2831
• External Method: seathru_nerf by @psetinek in #2829
• Add support for loading 3D points with ODM data processor by @pierotofy in #2838
• Add average_init_density to improve robustness of nerfacto training by @jb-ye in #2834
• Mask flexibility for methods than need to sample in the masked region by @ethanweber in #2647
• Add flag to disable apperance embedding by @jkulhanek in #2843
• Fix #2843 - safer typing by @jkulhanek in #2848
• limit torch to <=2.1 to avoid type errors by @kerrj in #2857
• Have predicted depth as expected_depth so gradients can be computed by @mattstrong-stanford in #2852
• Avoid OOM for caching undistorted images of splatfacto by @jb-ye in #2855
• Replace depth alpha mask logic with torch.where for gradients in splatfacto by @kerrj in #2856
• Changed torch version to <2.2 by @simonbethke in #2861
• Fix ODM data processor coordinate system by @pierotofy in #2847
• Upper-bound torch version for dev installs only by @brentyi in #2864
• v1.0.1 by @tancik in #2865

New Contributors

• @Harr7y made their first contribution in #2804
• @pyetras made their first contribution in #2825
• @AntonioMacaronio made their first contribution in #2823
• @psetinek made their first contribution in #2829
• @mattstrong-stanford made their first contribution in #2852
• @simonbethke made their first contribution in #2861

Full Changelog: v1.0.0...v1.0.1

v1.0.0

Highlights

• New splatfacto model, built on the gsplat implementation of INRIA's 2023 Gaussian Splatting algorithm.
• Complete rewrite of the viewer, now built entirely on viser.
• Project Aria and EyefulTower dataset support, and COLMAP pipeline improvements.

...and much more, thanks to the 60+ contributors who helped make this happen! See below.

What's Changed

• Add background_color config to keep consistent with other models by @kobejean in #2455
• Do not materialize mlp_head if no field_heads by @Mxbonn in #2457
• .vscode settings match move to ms-python.black-formatter by @Mxbonn in #2456
• Migrate doc from RTD to gh-pages by @liruilong940607 in #2460
• Allow custom sampler in ProposalNetworkSampler by @Mxbonn in #2464
• Fix COLMAP dataparser masks and depthmaps by @jkulhanek in #2467
• Track all viser GUI folders to avoid folder duplicates (viewer-beta) by @chungmin99 in #2466
• Update colmap installation instructions with conda by @ginazhouhuiwu in #2473
• Update doc.yml by @liruilong940607 in #2479
• Remove rtd by @tancik in #2478
• Don't deploy docs from pull requests by @brentyi in #2482
• fix:correction of PairPixelSampler in case of using masks with depth-nerfacto by @pierremerriaux-leddartech in #2477
• ns-render - render all images by @jkulhanek in #2459
• fixing broken links in README by @ARK3r in #2483
• Fourier Feature encodings and polyhedron encodings by @Mxbonn in #2463
• fixed link to viser to be functional link cause previous 404ed by @THE-COB in #2486
• Add support of using both Comet and Viewer at the same time by @Lothiraldan in #2488
• Fix DatasetRender to override dataset path only when provided by @sjj118 in #2494
• Remove unused config param by @tancik in #2495
• (minor) Align upsampling in TensorCPEncoding with TensorVMEncoding by @Dwawayu in #2493
• Add documentation about unknown depth pixel convention by @jampekka in #2490
• patch for mkdirs on camera path json filepath by @kerrj in #2502
• Move camera optimization out of datamanager and parallelize dataloading by @kerrj in #2092
• minor patch for exporting pcs and meshes with parallel datamanager by @kerrj in #2503
• Fix jitter viewer by @kerrj in #2505
• Add ODM data processor by @pierotofy in #2517
• Upgrade pyright by @brentyi in #2526
• Update URLs for readme and image references by @cvachha in #2527
• Add camera_type dropdown in new viewer by @cvachha in #2525
• Fix datamanagers' get_rays_per_batch functions in case of using DynamicBatchPipeline by @sjj118 in #2522
• Bump tyro to 0.5.10 by @brentyi in #2393
• Update data_conventions.md by @SeaOtocinclus in #2534
• Fix pillow version to avoid error with ANTIALIAS (Fix #2378) by @ahmed-shariff in #2446
• Fix multiclient jitter with separate render threads for each client by @ginazhouhuiwu in #2519
• Add scannetpp dataparser by @liu115 in #2498
• minor typing error by @Shiva-Gantha in #2545
• Missing a space in base_datamanager.py by @OrangeSodahub in #2546
• Fix pixel-perfect-sfm installation in Dockerfile by @2b-t in #2549
• static imgs never rendered in multiclient by @kerrj in #2551
• Move only per-chunk data to the GPU to support rendering large images by @decrispell in #2481
• make refine intrinsics during ns-process-data optional by @machenmusik in #2550
• Few viewer changes by @cvachha in #2553
• ns-process-data images fix: multinerf uses equals sign by @machenmusik in #2558
• Add gradient scaling option to more methods by @kobejean in #2555
• Expose camera frustums size and occlusions to CLI by @ethanweber in #2556
• Make resetting the near_plane in NearFarCollider optional. by @Mxbonn in #2465
• Print beta URL in banner correctly by @kerrj in #2569
• Fix torch.compile for torch 2.1 by @jkulhanek in #2577
• Remove compile line in camera_utils by @ginazhouhuiwu in #2584
• Fix HTML render of README.md by @krahets in #2599
• Render nearest training view by @AdamRashid96 in #2384
• Make spherical harmonics test more forgiving by @brentyi in #2607
• Update LICENSE by @cpheinrich in #2597
• Remove an outdated comment in RayGenerator class by @emmaguo13 in #2606
• Added UI to load camera path from existing path for the dataset by @THE-COB in #2530
• Bugfix; scrape pipeline for viewercontrol, not trainer by @chungmin99 in #2621
• Remove PCA colormap bug for float output maps by @chungmin99 in #2620
• Fix tangential camera distortion (p1,p2 in opencv camera) by @jkulhanek in #2627
• Integrate scene click capability from original viewer by @chungmin99 in #2509
• Keyframe timing for beta viewer by @brentyi in #2638
• Patch spiral rendering w/ parallel datamanager by @kerrj in #2637
• Bugfix: shape confusion in resizing by @Yosshi999 in #2565
• Make nuScenes dataparser compatible with ParallelDataManager by @dmholtz in #2635
• fix include_input for NeRFEncoding by @blacksino in #2642
• Converted many @DataClass assignments for python 3.11 compatibility by @jaggzh in #2630
• Mps fix by @Jerry-Master in #2436
• Fixes bug in PairPixelSampler when working with masked dataset by @padeler in #2368
• Encoder + MLP combo by @ethanweber in #2063
• Ns-process-data bug with dev version of colmap by @AdamRashid96 in #2651
• Fix typos by @omahs in #2655
• Update hloc and add new matcher types by @alexandrukis in #2658
• Trainer can now always save dataparser transform by @pwais in #2653
• Update dependencies for M1 Macs by @brentyi in #2665
• Update cameras.py - removed duplicate lines by @IamMohitM in #2661
• Fix large dataset (> 512 images) training in viewer_beta by @brentyi in #2669
• Add support for Project Aria datasets. by @sweeneychris in #2617
• Fix custom data links in README by @brentyi in https:...