v0.1.17

What's Changed

• Fix normal orientation loss by @tancik in #1298
• Tancik/protobuf version req update by @tancik in #1303
• Corrected the order of texture arguments by @MaroonAmor in #1300
• Add sphere collider by @tancik in #1310
• Explicitly require NVIDIA graphics cards in README by @ansonkao in #1313
• nerfplayer with ngp and nerfacto by @lsongx in #1277
• adding check if camera_path_dir exists by @terrancewang in #1319
• Fix crop robustness by @tancik in #1321
• bump nerfacc to 0.3.4 by @machenmusik in #1322
• Document localhost URL (fixes #1323) by @machenmusik in #1324
• Add Stable Diffusion helper code by @tancik in #1320
• [Fix] Docker not supporting older CUDA architectures by @Zunhammer in #1328
• Cleanup aabb intersection logic by @tancik in #1329
• Avoid protobuf 3.20.0 by @tancik in #1330
• Safe normalize by @tancik in #1332
• Add support for gradient clipping by @hturki in #1331
• Update optimizer docstrings by @tancik in #1333
• Dont move viewer outputs to cpu by @tancik in #1335
• add option to log training gradients by @hturki in #1334
• Check if curl is installed by @tancik in #1337
• Set min render value for viewer by @tancik in #1339
• Add lambertian renderer by @terrancewang in #1336
• Fix normals visualization by @tancik in #1341
• Fix depth-supervised nerf when using z-buffer depth. by @jeffreyhuparallel in #1340
• Create shaders by @tancik in #1344
• New config parameter --viewer.codec defaults to VP8, but allows H264 by @machenmusik in #1343
• double-quote filenames in process-data commands so that on Windows, paths with spaces don't fail by @machenmusik in #1347
• Improve viewer colormap choices by @tancik in #1348
• Consistent naming for depth maps in Polycam by @ManuConcepBrito in #1352
• Handle uncalibrated Metashape sensor by @gilureta in #1359
• add viewer.local config property when viewing local instance by @machenmusik in #1358
• Allow methods configs to be discovered from external packages. by @jkulhanek in #1357
• Trace Stable Diffusion by @tancik in #1362
• Fix ns-train help by @jkulhanek in #1370
• Improve ns-process-data images when using existing COLMAP model by @jkulhanek in #1371
• Add support for sparse depth maps by @yimingzhou1 in #1373
• Fix Multiple Typos by @iamNCJ in #1376
• Colmap docs note by @tancik in #1380
• add time by @lsongx in #1391
• Add support for is_fisheye flag in ingp transforms.json by @jkulhanek in #1385
• Add ns-viewer command to run viewer only with model in eval mode by @jkulhanek in #1379
• Update citation by @tancik in #1392
• Fix IPython error display by @M-Colley in #1397
• Update reality capture doc by @tancik in #1404
• sdfstudio dataparser/dataset by @pablovela5620 in #1381
• Downgrade nerfacc to 0.3.3 by @tancik in #1408
• v0.1.17 by @tancik in #1409

New Contributors

• @MaroonAmor made their first contribution in #1300
• @ansonkao made their first contribution in #1313
• @jeffreyhuparallel made their first contribution in #1340
• @gilureta made their first contribution in #1359
• @jkulhanek made their first contribution in #1357
• @yimingzhou1 made their first contribution in #1373
• @iamNCJ made their first contribution in #1376
• @M-Colley made their first contribution in #1397
• @pablovela5620 made their first contribution in #1381

Full Changelog: v0.1.16...v0.1.17

v0.1.16

Highlights

• Blender VFX add-on
• Scene cropping
• Basic RealityCapture Support
• Improved camera path management
• Bug fixes

What's Changed

• Bump json5 from 1.0.1 to 1.0.2 in /nerfstudio/viewer/app by @dependabot in #1238
• Loosen dependencies by @tancik in #1240
• Docs fix by @Mason-McGough in #1244
• create video writer later, to adapt to render image's width and height (fixes #1242) by @machenmusik in #1245
• Update transform.json docs by @tancik in #1246
• Bump nerfacc to 0.3.3 by @tancik in #1249
• setting tensorf default from wandb to viewer by @terrancewang in #1250
• Rollback cryptography to fully working version (pre-39; 38 works) by @machenmusik in #1254
• NeRFPlayer sliding window implementation by @lsongx in #1255
• downsample all images by @machenmusik in #1259
• Camera path by @THE-COB in #1251
• CropBox GUI implementation by @dozeri83 in #1205
• use -noautorotate option to avoid discrepancies (fixes #1260) by @machenmusik in #1261
• Add support for RealityCapture by @tancik in #1055
• Blender addon script and documentation by @cvachha in #1218
• Fix tensorf chkpt loading by @terrancewang in #1266
• Bug fix for loading cameras with different distortions in nerfstudio_dataparser by @matsuren in #1268
• refactor control panel by @tancik in #1267
• Add step argument in eval_utils.py to fix rendering bug. by @Mason-McGough in #1272
• DyCheck dataset parser by @lsongx in #1270
• Increase max_res of first proposal network by @kerrj in #1282
• Fix total train time metric logging by @williamshen-nz in #1284
• Depth data loading for Polycam LiDAR data by @ManuConcepBrito in #1253
• Fix the issue with missing images in the Reality Capture data. by @dkorolov in #1285
• Allow users to modify background color when cropping by @tancik in #1289
• Updated Blender add-on calculating duration by @cvachha in #1288
• Remove bracket from ffmpeg command's crop filter by @heymarcell in #1294
• Enable rendering with crop by @tancik in #1295
• Load path dropdown by @THE-COB in #1290
• PIL Dependency by @mcallisterdavid in #1197
• v0.1.16 by @tancik in #1296

New Contributors

• @lsongx made their first contribution in #1255
• @cvachha made their first contribution in #1218
• @williamshen-nz made their first contribution in #1284
• @ManuConcepBrito made their first contribution in #1253
• @dkorolov made their first contribution in #1285
• @heymarcell made their first contribution in #1294

Full Changelog: v0.1.15...v0.1.16

v0.1.15

Highlights

• Equirectangular output now have metadata for rendering 360 on platforms like youtube
• Support grayscale images
• Update nerfacc to 0.3.2
• New depth supervised nerfacto model (run with depth-nerfacto)
• Various bug fixes

What's Changed

• [Fix] point cloud export broken in Dockerfile by @Zunhammer in #1148
• when camera_type is equirectangular, add spherical metadata to rendered video by @machenmusik in #1151
• work around mismatch to allow rgb depth side-by-side output by @machenmusik in #1156
• Fix dynamic batch logic during inference by @tancik in #1160
• fix typo: "dictioary" -> "dictionary" by @JamesPerlman in #1162
• Support grayscale-only images in dataset by @mabl in #1142
• PixelSampler refactor by @akristoffersen in #1163
• Fix downscaling and type errors by @tancik in #1164
• Update docstrings by @tancik in #1165
• Type and docs fixes by @brentyi in #1170
• Force 8 bit output from ffmpeg by @tancik in #1180
• Qt5 hotfix, should fix ns-export by @akristoffersen in #1181
• TrainerConfig inherit from InstantiateConfig by @jake-austin in #1175
• Created using Colaboratory by @tancik in #1183
• Fix typo in mlp.py by @eltociear in #1185
• upgrade nerfacc to 0.3.x with CUB accel by @liruilong940607 in #1187
• Quick fix for mask None check by @ethanweber in #1193
• Camera export by @mackopes in #1191
• Making camera selection on viewer persistent by @THE-COB in #1192
• Add hashmap size and num_levels to instant_ngp_field, issue 1158 by @DreekFire in #1196
• fixed an exception bug in case of num_samples is 0 by @dozeri83 in #1198
• Depth supervision for nerfacto by @mpmisko in #1173
• Added missing arg to ns-download-data command by @iamrajee in #1201
• Render cameras with render times for dynamic nerfs by @nlml in #1199
• Added ETA to base config for the event writer by @THE-COB in #1203
• fix notebook, box drawing by @3a1b2c3 in #1202
• zqm linting by @tancik in #1233
• Add mask conventions to docs by @tancik in #1234
• Fix typo by @matsuren in #1231
• Update incorrect comment by @JamesPerlman in #1226
• Update Quickstart instructions to remove '--dataset' (#1222) by @Mason-McGough in #1230
• Non-strict loading of checkpoints by @tancik in #1235
• save image_coords as non-persistent buffer by @hturki in #1216
• Add cropping options to data processing by @tancik in #1236
• v0.1.15 by @tancik in #1237

New Contributors

• @JamesPerlman made their first contribution in #1162
• @mabl made their first contribution in #1142
• @eltociear made their first contribution in #1185
• @THE-COB made their first contribution in #1192
• @DreekFire made their first contribution in #1196
• @mpmisko made their first contribution in #1173
• @iamrajee made their first contribution in #1201
• @nlml made their first contribution in #1199
• @3a1b2c3 made their first contribution in #1202
• @matsuren made their first contribution in #1231
• @Mason-McGough made their first contribution in #1230

Full Changelog: v0.1.14...v0.1.15

v0.1.14

Highlights

• Additional export video camera options (fisheye, equirectangular)
• KIRI Engine support (Android/IOS capture app)
• Speedups
• Bug Fixes

What's Changed

• Colab working with Python3.8 by @akristoffersen in #1109
• Fix circular imports by @nikmo33 in #1115
• Fix slow training issue when using mask by @ttaoREtw in #1114
• change normals grad calculation by @terrancewang in #1108
• Log start iter by @tancik in #1119
• instant-ngp-bounded model by @dozeri83 in #1116
• Missing images error warning by @tancik in #1122
• Fix datamanager dataparser_outputs by @tancik in #1123
• Add docs for KIRI Engine by @tancik in #1124
• Changed tinycudann to wheel to work with py3.8 by @akristoffersen in #1126
• Immediately update viewer when output or cmap changes by @tancik in #1131
• Use torch.where for faster contract by @jenkspt in #1132
• Update render ui to use leva by @tancik in #1133
• Update Render View logic by @tancik in #1134
• adding data subset by @ethanweber in #1135
• Minimal disk dataparser by @nikmo33 in #1128
• Upgraded Gdown to fix data downloading. by @akristoffersen in #1140
• Support non-pinhole rendering by @tancik in #1141
• Fix Discord link in the docs by @mackopes in #1143
• v0.1.14 by @tancik in #1145

New Contributors

• @ttaoREtw made their first contribution in #1114
• @jenkspt made their first contribution in #1132
• @mackopes made their first contribution in #1143

Full Changelog: v0.1.13...v0.1.14

v0.1.13

Highlights

• Support for hloc toolbox in ns-process-data
• nerfacto quality improvements
• Bugfixes

What's Changed

• WIP : support for hloc toolbox (superpoint, superglue, netvlad,...) by @RandomPrototypes in #1026
• Metashape parsing fix by @tancik in #1084
• 'get docker' typo by @sandros94 in #1089
• Update mipnerf defaults by @ethanweber in #1092
• Use max resolution when not moving or training by @machenmusik in #1091
• Fix mask filenames misaligned after splitting set into train and eval by @Zunhammer in #1093
• Fix crash on camera delete by @tancik in #1094
• Fix IndexError if ther is no mask filenames by @ksnzh in #1095
• Bug correction auto_orient_and_center_poses by @fdarmon in #1098
• Save out dataparser transforms by @tancik in #1105
• Bump max res default by @tancik in #1106
• v0.1.13 by @tancik in #1107

New Contributors

• @RandomPrototypes made their first contribution in #1026
• @sandros94 made their first contribution in #1089
• @fdarmon made their first contribution in #1098

Full Changelog: v0.1.12...v0.1.13

v0.1.12

Highlights

• Added mesh export
• Added per-frame intrinsics for nerfstudio and polycam data
• Added support for equirectangular images
• Various bug fixes and improvements

What's Changed

• Check is metashape xml is present by @tancik in #1044
• add render time in viewer by @JulianKnodt in #1007
• Add Support for Equirectangular Images by @kevinddchen in #966
• Add per frame intrinsic support to nerfstudio data by @tancik in #1049
• Add per frame polycam intrinsics by @tancik in #1050
• Quick fix to viewer OOM issue by @ethanweber in #1053
• Fix torch spherical harmonics by @kevinddchen in #1054
• Fix metashape labels by @tancik in #1060
• Fix IndexError runtime error by @machenmusik in #1057
• Make friends_dataparser path separator agnostic (fixes #1066) by @machenmusik in #1067
• Fix device mismatch error when using masks by @hturki in #1061
• Dnerf viewer and training fix by @myaldiz in #1065
• More informative completion install error messages by @brentyi in #1068
• Interpolation config for tcnn hashgrid by @ydmalex in #1058
• limit semantic-nerfw iterations to 30000 by @machenmusik in #1074
• Mesh export by @ethanweber in #809
• [WIP] Mesh export UI by @ethanweber in #1076
• v0.1.12 by @ethanweber in #1079

New Contributors

• @machenmusik made their first contribution in #1057
• @hturki made their first contribution in #1061
• @myaldiz made their first contribution in #1065
• @ydmalex made their first contribution in #1058

Full Changelog: v0.1.11...v0.1.12

v0.1.11

Highlights

• Add support for Metashape poses
• Add normals prediction for the nerfacto model
• Add docker support
• Data loading speedups
• Improve viewer connection stability (if you are having new issues, add --viewer.skip-openrelay to command and let us know)
• Core now support cameras with variable intrinsics

What's Changed

• Update viewer sensitivity by @tancik in #975
• Refactor process data by @tancik in #976
• Simplify readme by @tancik in #982
• Support downsample in dataset by @nikmo33 in #973
• Computing and predicting normals by @ethanweber in #905
• Add Metashape support by @tancik in #989
• fix typo error which should be self.cameras.times by @ksnzh in #991
• TensoRF bugfix: AABB normalized_positions to [-1, 1] by @JulianKnodt in #998
• Simplify documentation by @tancik in #1003
• Add metashape to docs by @tancik in #1004
• create rays on device when rendering by @zmurez in #993
• implemented threadpool mechanics when loading images by @dozeri83 in #994
• (minor) --trainer.load_dir => --trainer.load-dir in docs by @brentyi in #1010
• add support for per-image mask filenames in transforms.json by @decrispell in #957
• Fix dataparser filename bug by @tancik in #1013
• Improve webrtc stability by @tancik in #1014
• Handle missing metashape data by @tancik in #1015
• Updating the script for downloading data by @jake-austin in #968
• Cameras TensorDataclass and Phototourism by @ethanweber in #961
• Typo in ProcessRecord3D error by @akristoffersen in #1027
• config nerfacto by @zmurez in #1020
• Pinched Looping Camera Path Fix by @mcallisterdavid in #1029
• Cleanup train cli logs by @tancik in #1037
• Camera Label Hiding by @mcallisterdavid in #1031
• Ignore non-image files in ns-process-data by @tancik in #1040
• Add Docker support by @Zunhammer in #1022
• Add support for insta360 single-file format by @cnsumner in #1024
• Fix benchmarking scripts by @devernay in #1005
• Allow users to skip using openrelay by @tancik in #1042
• v0.1.11 by @tancik in #1043

New Contributors

• @ksnzh made their first contribution in #991
• @dozeri83 made their first contribution in #994
• @Zunhammer made their first contribution in #1022
• @cnsumner made their first contribution in #1024
• @devernay made their first contribution in #1005

Full Changelog: v0.1.10...v0.1.11

v0.1.10

TLDR

• Polycam data now supported, learn more here
• DNeRF added
• NuScenes data supported

What's Changed

• Update docs by @tancik in #944
• Update viewer zoom sensitivity by @tancik in #945
• Add flag to kill training when complete by @tancik in #948
• Dont crash on invalid websocket by @tancik in #949
• Update viewer docs by @tancik in #950
• Streamline Viewer in Colab, added more checks before training by @akristoffersen in #947
• NuScenes dataparser by @zmurez in #939
• Colab hotfix: forgot glob import by @akristoffersen in #952
• Remove unimplemented interp render option by @tancik in #953
• Filter torchvision warnings by @tancik in #954
• Add 'inference' mode to pipeline by @nikmo33 in #956
• Fix python 3.7 support by @tancik in #959
• Bump loader-utils from 2.0.2 to 2.0.4 in /nerfstudio/viewer/app by @dependabot in #962
• Add temporal distortion and D-NeRF model by @JulianKnodt in #825
• Small Typos by @akristoffersen in #960
• Cleanup configs by @tancik in #963
• Add support for polycam by @tancik in #958
• Fix loading in inference mode by @tancik in #965
• v0.1.10 by @tancik in #974

New Contributors

• @zmurez made their first contribution in #939
• @dependabot made their first contribution in #962

Full Changelog: v0.1.9...v0.1.10

v0.1.9

TLDR

• Improved animation controls
• Added the ability to animate focal length
• Point cloud exporting
• Bug fixes

What's Changed

• Add docs for nerfacto by @tancik in #891
• Camera and world coordinate conventions by @ethanweber in #889
• Start viewer with more accurate speed calculation by @tancik in #899
• Minor doc updates by @tancik in #903
• Refactor Data Pipeline by @nikmo33 in #896
• Use tan instead of tanh by @tancik in #913
• Terrance/camera controls by @terrancewang in #748
• Camera Controls Overhaul (FOV + Keyframing) by @origamiman72 in #792
• Add training finished statement by @tancik in #914
• fixing typos by @Mirmix in #915
• [WIP] nan-training-debug by @nikmo33 in #910
• Allow specifying colmap path by @tancik in #917
• 3D Viewer Displays cameras' Chosen Names by @mcallisterdavid in #920
• Fixed crash when camera looping is on, but no cameras by @origamiman72 in #922
• Support exporting point clouds by @tancik in #924
• Add viewport controls help modal by @tancik in #927
• Avoid boolean flag auto conversion by @tancik in #929
• Missing mlp output activation by @tancik in #930
• Do not apply activation for last MLP layer by @kevinddchen in #932
• Add missing activation by @tancik in #933
• Add missing init.py files by @brentyi in #935
• Record3D added to Colab by @akristoffersen in #926
• Add exporting docs by @tancik in #940
• Add pointcloud export panel to UI by @tancik in #938
• v0.1.9 by @tancik in #941

New Contributors

• @Mirmix made their first contribution in #915

Full Changelog: v0.1.8...v0.1.9

v0.1.8

TLDR

• Record3D users should use the ns-process-data script to convert their data to nerfstudio-data. The record3d-data option in ns-train will be deprecated.
• Mostly Bugfixes

What's Changed

• Fix ffmpeg call by @tancik in #856
• Fix camera interpolation bug by @tancik in #858
• Add colab notebook to repo by @tancik in #864
• Viewer visualization crashes when using DDP by @thomasw21 in #863
• Update colab links to use repo colab by @tancik in #866
• Add more debug logs to data processing by @tancik in #867
• Update readme by @tancik in #869
• Don't add relative paths to experiment name by @tancik in #871
• Add control for resampling images by @kevinddchen in #847
• always list paths as posix paths by @terrancewang in #860
• Check main thread when always rendering. by @thomasw21 in #865
• Add median depth calculation by @tancik in #877
• remove sf street by @terrancewang in #876
• disallow fars <= nears by @decrispell in #881
• add correct none check to distortion params by @nikmo33 in #878
• DNeRF Data Parser by @JulianKnodt in #817
• tensorf by @terrancewang in #586
• Update render UI by @tancik in #885
• [WIP] Record3D Dataparser moved to ns-process-data script. by @akristoffersen in #872
• v0.1.8 by @tancik in #890

New Contributors

• @thomasw21 made their first contribution in #863
• @decrispell made their first contribution in #881
• @nikmo33 made their first contribution in #878

Full Changelog: v0.1.7...v0.1.8