v0.3.3

Highlights

• Faster downscaling during image processing
• VR180 rendering
• nerfacto alpha transparency training
• Sparse-NeRF depth losses
• New template repo for implementing new methods
• Bug fixes

What's Changed

• Add examps to generfacto docs by @terrancewang in #2120
• Fix get_num_frames_in_video by @olegpokhilchenko in #2090
• Update nerfacto.md by @tancik in #2126
• Fix pickling of VanillaDataManager by @jkulhanek in #2129
• Jkulhanek/colmap dataparser by @jkulhanek in #2115
• Update generfacto.md by @tancik in #2132
• Fix to fall back to torch if tcnn not found by @ethanweber in #2124
• allow ns-process-data to use CR2 raw images by @machenmusik in #2119
• Fix unnecessary encodings calculation in tensorf field by @kst179 in #2145
• Remove breakpoint from colmap_dataparser by @jkulhanek in #2141
• Add option to render evaluated images by @hturki in #2135
• Fix typo in nerfacto.py by @MvWouden in #2149
• add build-arg cuda version to dockerfile by @Ilyabasharov in #2139
• Fix use_appearance_embedding for instant-ngp by @brentyi in #1999
• Default to tcnn by @tancik in #2159
• make all SDFFieldConfig params configurable via constructor by @decrispell in #2167
• Correct path notation for Windows in documentation by @KaerMorh in #2171
• Fix image saving format issue in render by @colin-de in #2172
• fix the ZeroDivisionError occured in int(image_width / aspect_ratio) when aspect_ratio equals zero by @WebGao in #2177
• Remove pytorch hash table if tcnn implementation is used by @georghess in #2175
• Feature/run_beta_viewer by @Ilyabasharov in #2157
• Avoid zero duration crash by @machenmusik in #2193
• Fixing Colmap VocabTreeMatching command to support spaces in Filename by @JDihlmann in #2195
• Default fied_heads with instances will cause shared parameters by @hyperzy in #2197
• Pin versions of pyceres+pixsfm, up pycolmap by @skydes in #2199
• Adding input handling for RFFEncoding's get_out_dim(self) by @CardiacMangoes in #2201
• Fix colmap vocab_tree_path (#2195) to work with Windows by @machenmusik in #2204
• Add saving/loading scheduler state_dict by @hyperzy in #2203
• Omnidirectional Stereo Update - Alignment and Compositing by @cvachha in #2217
• Reorient point cloud normals for better surface reconstruction by @NotAGoose in #2220
• Try non-strict model loading on RuntimeError. ... by @sjuxax in #2214
• Add documatations for TensoRF by @Xzzit in #2209
• Fast downscaling for video and image processing (PR2168 plus images) by @machenmusik in #2213
• refactor: get_normalized_directions by @SauravMaheshkar in #2216
• Update torchmetrics by @tancik in #2223
• Remove superfluous -r1 (fix ffmpeg/colmap video) by @SharkWipf in #2231
• Fix arguments that are not passed from config to NerfactoField by @liu115 in #2186
• Minor kwarg fix to be future-compatible with viser version by @kerrj in #2235
• Add std for eval by @Khoa-NT in #2239
• Implement nerfacto alpha transparency training by @nepfaff in #2165
• Load schedulers from ckpt in a backward-compatible manner by @hyperzy in #2236
• support packed samples in semantic renderer by @hturki in #2247
• Render VR180 video by @cvachha in #2261
• nerfbusters doc by @FrederikWarburg in #2208
• Use correct pixel area in volumetric sampler by @hturki in #2268
• ns-render with depth colormap support by @ethanweber in #2125
• Update index.md by @tancik in #2279
• Moved Volinga as an external method by @Frivas97 in #2282
• Add depth_unit_scale_factor to sdfstudio_dataparser by @alexsax in #2285
• Option to process images on GPU for speed at the expense of memory by @AX-I in #2288
• More robust install instructions by @brentyi in #2277
• Remove system-level CUDA install from install instructions by @brentyi in #2289
• Monodepth Regularization by @ethanweber in #2284
• Allow same dimensions optimizations to be disabled by @machenmusik in #2292
• convert raw to JPG not TIFF, to avoid fast-downscale ffmpeg error by @machenmusik in #2295
• Train and eval data splits (Nerfbusters) by @FrederikWarburg in #2207
• PixelSampler Fixes by @AdamRashid96 in #2298
• Add sponsors by @tancik in #2307
• Adding support for PairPixelSampler and PatchPixelSampler masks by @ethanweber in #2305
• template repository by @maturk in #2319
• When generating depth images, use torch_compile with ZoeDepth for compatibility by @machenmusik in #2303
• Rename num_cameras to camera_indices by @maturk in #2326
• Ddp docs by @yurkovak in #2332
• fix: Allow custom user ID by @yurkovak in #2334
• fix: ColmapDataParser downscaling images error. by @onpix in #2337
• fix #2318: add train/test/eval split for instant ngp dataparser by @elenacliu in #2335
• Circle crop to images processing pipeline by @kerrj in #2341
• viser 0.1.0, fix type checker errors by @brentyi in #2343
• Update pixel_samplers.py by @alexsax in #2327
• v0.3.3 by @tancik in #2349

New Contributors

• @olegpokhilchenko made their first contribution in #2090
• @kst179 made their first contribution in #2145
• @Ilyabasharov made their first contribution in #2139
• @KaerMorh made their first contribution in #2171
• @WebGao made their first contribution in #2177
• @georghess made their first contribution in #2175
• @JDihlmann made their first contribution in #2195
• @hyperzy made their first contribution in #2197
• @skydes made their first contribution in #2199
• @CardiacMangoes made their first contribution in #2201
• @NotAGoose made their first contribution in #2220
• @sjuxax made their first contribution in #2214
• @Xzzit made their first contribution in #2209
• @liu115 made their first contribution in #2186
• @FrederikWarburg made their first contribution in #2208
• @alexsax made their first contribution in #2285
• @AX-I made their first contribution in #2288
• @AdamRashid96 made their first contribution in #2298
• @yurkovak made their first contribution in #2332
• @onpix made their first contribution in #2337
• @elenacliu made their first contribution in #2335

Full Changelog: v0.3.2...v0.3.3