openftgs 0.1.0__tar.gz

openftgs-0.1.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

openftgs-0.1.0/NOTICE

@@ -0,0 +1,84 @@
+openftgs
+Copyright 2026 Samuel Austin
+
+
+Third-party software
+--------------------
+
+openftgs includes a copy of RoMa's inference code in the package openftgs._vendor.romatch
+(folder openftgs/_vendor/romatch). The license texts named below sit in that folder: RoMa's
+as LICENSE, the others in its LICENSES folder. Its VENDOR.json records the upstream commit,
+the SHA-256 of every file and license text, and how each file was changed; every changed file
+carries a modification notice.
+
+RoMa (https://github.com/Parskatt/RoMa), commit 77f8d68803526dcddfd9b7a46bc76125bdc25f15
+Copyright (c) 2023 Johan Edstedt
+Licensed under the MIT License (LICENSE).
+
+The DINOv2 layers in RoMa's models/transformer (https://github.com/facebookresearch/dinov2)
+Copyright (c) Meta Platforms, Inc. and affiliates.
+Licensed under the Apache License, Version 2.0 (LICENSES/DINOv2.txt). DINOv2 has no NOTICE
+file. Their file headers reference:
+- DINO (https://github.com/facebookresearch/dino), licensed under the Apache License,
+  Version 2.0 (LICENSES/DINO.txt);
+- timm (https://github.com/huggingface/pytorch-image-models), Copyright 2019 Ross Wightman,
+  licensed under the Apache License, Version 2.0 (LICENSES/timm.txt).
+
+Patch2Pix (https://github.com/GrumpyZhou/patch2pix), credited in RoMa's utils/utils.py
+Copyright (c) 2020 Qunjie Zhou
+Licensed under the MIT License (LICENSES/Patch2Pix.txt).
+
+LoFTR (https://github.com/zju3dv/LoFTR), from which RoMa's utils/utils.py adapts code
+Licensed under the Apache License, Version 2.0 (LICENSES/LoFTR.txt).
+
+Kornia (https://github.com/kornia/kornia), named as a precaution for the epipolar-distance
+helpers in RoMa's utils/utils.py, which import kornia
+Licensed under the Apache License, Version 2.0 (LICENSES/Kornia.txt).
+
+InLoc (https://github.com/HajimeTaira/InLoc_demo), whose approach RoMa's utils/utils.py credits
+Copyright (c) 2017 Hajime Taira
+Licensed under the MIT License (LICENSES/InLoc.txt).
+
+openftgs includes a copy of NVIDIA's ꟻLIP loss in the module openftgs._vendor.flip.flip_loss
+(folder openftgs/_vendor/flip), with changes that the file marks and its VENDOR.json lists, and
+its tests include the unmodified file as a reference (folder tests/reference/flip). Both folders
+hold ꟻLIP's license as LICENSE.
+
+ꟻLIP (https://github.com/NVlabs/flip), commit d3ab02a4425eb167d8a4021b382a2bafc6e6c838,
+src/pytorch/flip_loss.py
+Copyright (c) 2020-2025, NVIDIA CORPORATION & AFFILIATES (the file's own header reads
+2020-2024). All rights reserved.
+Licensed under the BSD 3-Clause License (LICENSE).
+
+openftgs adapts piq's HaarPSI in the module openftgs.losses (the functions _haarpsi_constants,
+_similarity and _haarpsi), with the changes its comments list.
+
+piq (https://github.com/photosynthesis-team/piq), version 0.8.0, piq/haarpsi.py and the helpers
+it calls (piq/functional/colour_conversion.py, piq/functional/filters.py, piq/functional/base.py)
+By Sergey Kastryulin and the piq contributors (the package's metadata names its author; its
+license file names no copyright holder, and it has no NOTICE file).
+Licensed under the Apache License, Version 2.0 (https://www.apache.org/licenses/LICENSE-2.0).
+
+The documentation site's template docs/_templates/autosummary/module.rst (in the source
+repository only; neither the sdist nor the wheel holds it) follows the layout of Sphinx's
+autosummary/module.rst; its header comment carries Sphinx's license in full.
+
+Sphinx (https://www.sphinx-doc.org/), version 9.2.0,
+sphinx/ext/autosummary/templates/autosummary/module.rst
+Copyright (c) 2007-2026 by the Sphinx team (see AUTHORS file). All rights reserved.
+Licensed under the BSD 2-Clause License.
+
+
+Pretrained weights
+------------------
+
+openftgs downloads the pretrained weights of its trunk perceptual networks from their
+publishers at first use; it neither includes nor redistributes them.
+
+The perceptual network vgg16_oxford uses the 16-layer VGG model of the Visual Geometry Group,
+University of Oxford, from its Caffe release
+(https://www.robots.ox.ac.uk/~vgg/research/very_deep/), whose convolutions openftgs converts
+into the user's torch.hub cache at first use. The model is licensed under the Creative Commons
+Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/), with
+attribution to its technical report: Karen Simonyan and Andrew Zisserman, "Very Deep
+Convolutional Networks for Large-Scale Image Recognition", arXiv:1409.1556 (2014).

openftgs-0.1.0/PKG-INFO

@@ -0,0 +1,143 @@
+Metadata-Version: 2.4
+Name: openftgs
+Version: 0.1.0
+Summary: An independent clean-room implementation of FreeTimeGS (Wang et al., CVPR 2025), not affiliated with its authors.
+Author: Samuel Austin
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/samaust/openftgs
+Project-URL: Source, https://github.com/samaust/openftgs
+Project-URL: Documentation, https://openftgs.readthedocs.io
+Project-URL: Issues, https://github.com/samaust/openftgs/issues
+Project-URL: Changelog, https://github.com/samaust/openftgs/blob/main/docs/changelog.md
+Keywords: gaussian-splatting,dynamic-scene-reconstruction,novel-view-synthesis,multi-view-video,4d-reconstruction,pytorch
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: GPU :: NVIDIA CUDA
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Image Processing
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Requires-Dist: gsplat==1.5.3
+Requires-Dist: torch>=2.1
+Requires-Dist: safetensors>=0.4
+Requires-Dist: PyYAML>=6.0
+Requires-Dist: Pillow>=10
+Requires-Dist: scipy>=1.10
+Requires-Dist: numpy>=1.23.2
+Provides-Extra: init
+Requires-Dist: torch>=2.5.1; extra == "init"
+Requires-Dist: torchvision; extra == "init"
+Requires-Dist: einops; extra == "init"
+Requires-Dist: loguru; extra == "init"
+Requires-Dist: tqdm>=4.70.1; extra == "init"
+Provides-Extra: fast
+Provides-Extra: dev
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-xdist; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
+Requires-Dist: hypothesis; extra == "dev"
+Requires-Dist: ruff; extra == "dev"
+Requires-Dist: mypy; extra == "dev"
+Requires-Dist: pip-licenses; extra == "dev"
+Requires-Dist: scikit-image; extra == "dev"
+Requires-Dist: opencv-python-headless; extra == "dev"
+Dynamic: license-file
+
+# openftgs
+
+[![CI](https://github.com/samaust/openftgs/actions/workflows/ci.yml/badge.svg)](https://github.com/samaust/openftgs/actions/workflows/ci.yml) [![Coverage](https://raw.githubusercontent.com/samaust/openftgs/main/docs/coverage.svg)](#tests) [![Documentation](https://app.readthedocs.org/projects/openftgs/badge/)](https://openftgs.readthedocs.io)
+
+openftgs is an independent clean-room implementation of FreeTimeGS (Wang et al., CVPR 2025), not affiliated with its authors.
+
+Python package `openftgs`, under the Apache-2.0 license, with the command-line tool `ftgs`.
+
+The documentation is at [openftgs.readthedocs.io](https://openftgs.readthedocs.io): installation, a quickstart, the datasets, the method, the command line and the API reference.
+
+**Status: alpha, version 0.1.0.** The first release, at the end of the plan's milestones M0 to M18: the library and its command line are complete, the differential harness checks them, and the method is validated on openftgs-synth, the owner's CC0-1.0 rendered dataset, rather than on the paper's datasets, so no result is compared with the paper's tables. The [changelog](https://github.com/samaust/openftgs/blob/main/docs/changelog.md) lists what the release contains and its known limitations.
+
+## Clean-room process
+
+The code is written from the paper (arXiv 2506.05348 v2) and from the owner's specification, which was itself written from the paper alone. No code, notebook, checkpoint or issue thread from any other implementation of the method, including the authors' own, is consulted. Third-party software is read only within its license: source code only under permissive licenses that allow commercial use, documentation only for everything else. Every source consulted is logged. The renderer the paper's project page links was read for a documentation-only survey ([#352](https://github.com/samaust/openftgs/issues/352)), and nothing from it is in the package.
+
+- [`docs/spec/openftgs-spec.md`](https://github.com/samaust/openftgs/blob/main/docs/spec/openftgs-spec.md): the specification, the source of truth for the implementation.
+- [`docs/clean-room/source-log.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/source-log.md): every source consulted during the implementation and what was taken from it.
+- [`docs/clean-room/questions.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/questions.md): gaps found in the spec and the provisional decisions made for them. GitHub doesn't render a Markdown file of this size, so it may show only the source; the [raw file](https://github.com/samaust/openftgs/raw/main/docs/clean-room/questions.md) opens as plain text.
+- [`docs/clean-room/milestones.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/milestones.md): the order in which the package was built.
+- [`docs/clean-room/release-checklist.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/release-checklist.md): the release-time checks of this version: the license scan, the libraries bundled in dependency wheels, the vendored code's license files, the pretrained weights' terms, the absence of dataset material, and the source-log audit.
+
+## Installation
+
+```bash
+pip install openftgs
+```
+
+openftgs needs Python 3.11 or later. Training needs a CUDA build of PyTorch. On Linux x86_64 the default `torch` wheel on PyPI already is one (torch 2.14.1 bundles the CUDA 13.0 libraries), so the command above installs it. For another CUDA version, or on Windows, install torch from the [PyTorch index](https://pytorch.org/get-started/locally/) before installing openftgs. A CPU build is enough for the tests on the reference backend; CI uses one because its runners have no GPU.
+
+Optional extras:
+
+- `[init]`: the dependencies of the RoMa-based 4D initialization (torchvision, einops and loguru, and tqdm for the code of XFeat that Tiny RoMa loads, Q-276), and of the four DINO perceptual networks (`dinov2_s14`, `dinov2_b14`, `dino_s16` and `dino_b16` as `loss.perc_net`), which run on the vendored RoMa's DINOv2 layers: without the extra, building one is an error that names it (Q-248). ꟻLIP, the default perceptual network, HaarPSI and the two Caffe trunks (`vgg16_oxford`, `alexnet_bvlc`) need no extra. RoMa's optional speed-up, [fused-local-corr](https://github.com/Parskatt/fused-local-corr) (MIT), isn't part of it: every release since 0.3.27 requires one exact torch release (0.3.211 requires torch 2.11.0), so installing one would replace your torch, and the newest unpinned wheel, 0.2.4, built for CUDA 12, doesn't load with torch 2.14.1 (D-85). On Linux, build it from source against the torch you have; this needs the CUDA toolkit (`nvcc`) and a CUDA build of PyTorch already installed:
+
+  ```bash
+  pip install --no-build-isolation git+https://github.com/Parskatt/fused-local-corr@689f2f7303ca0d97ac1f3d1364c732c6d8e889a8
+  ```
+
+  Without it, RoMa runs with its `use_custom_corr` option off and computes the local correlation in plain PyTorch, which is slower.
+- `[fast]`: currently empty. The faster SSIM loss uses [fused-ssim](https://github.com/rahul-goel/fused-ssim) (MIT), which is not on PyPI. Install it from its repository, at the commit openftgs is tested with; this needs the CUDA toolkit (`nvcc`) and a CUDA build of PyTorch already installed:
+
+  ```bash
+  pip install --no-build-isolation git+https://github.com/rahul-goel/fused-ssim@a7c48d6dd7ac6dc39a7958c7c4452e0b10418f38
+  ```
+
+  Without it, the SSIM loss runs in plain PyTorch and gives the same values.
+
+For development, install from a clone of the repository, with the test and lint tools:
+
+```bash
+pip install -e .[dev]
+```
+
+The differential harness is a separate package in `harness/` (`pip install -e harness`), which isn't published.
+
+## Command line
+
+```bash
+ftgs convert n3dv SRC OUT                        # a dataset into a scene manifest (needs ffmpeg)
+ftgs cache OUT/scene.json                        # decode the training frames once (optional)
+ftgs init roma OUT/scene.json init.safetensors   # initial points with RoMa ([init], a CUDA device)
+ftgs train OUT/scene.json --init init.safetensors --out RUN_DIR [--config CFG.json] [--seed S]
+ftgs train OUT/scene.json --init init.safetensors --out RUN_DIR --resume RUN_DIR/checkpoint
+ftgs export RUN_DIR/checkpoint --format ftgs-if --out model.safetensors
+ftgs render model.safetensors --request request.json --out renders.safetensors
+ftgs eval RUN_DIR/checkpoint OUT/scene.json --split test --out METRICS.json
+```
+
+`ftgs <command> --help` lists every option. A run directory holds `config.json`, `train_log.jsonl`, `checkpoint/` and, when the scene has a test split, `METRICS.json`; Ctrl-C saves the checkpoint before stopping. Exit code 0 is success, 2 an invalid input or environment, 3 a numerical failure. See [`docs/modules/cli/command-line.md`](https://github.com/samaust/openftgs/blob/main/docs/modules/cli/command-line.md).
+
+## Tests
+
+```bash
+pytest                       # CPU tests; GPU, dataset and weights tests skip when their resource is missing
+pytest -n auto --dist loadgroup -m "not gpu"      # the CPU tests on every physical core (Q-214)
+pytest -n auto --dist loadgroup -m "not gpu" --cov --cov-report=html   # with coverage, in htmlcov/ (Q-216)
+pytest -m gpu                # tests that need a CUDA device
+OPENFTGS_DATASETS=/path pytest -m dataset         # converter checks on real datasets
+OPENFTGS_ALLOW_DOWNLOADS=1 pytest -m weights      # tests that download pretrained weights
+python scripts/license_scan.py                    # GPL, AGPL, non-commercial fail; others need a sign-off in scripts/license_signoff.json
+```
+
+The coverage badge is the CPU and GPU suites' total on a machine with a CUDA device, the dataset and weights tests left out. `python scripts/coverage_local.py` runs both suites, writes `htmlcov/` and `coverage.xml`, and rewrites `docs/coverage.svg` when both pass; commit the SVG to update the badge (Q-217). CI's `test` job runs the CPU suite on Python 3.14 and on 3.11, the oldest Python openftgs supports (Q-1), and reports its coverage in its summary (Q-216).
+
+The license scan passes in CI's environment (CPU torch) and, for this release, in an environment with CUDA torch, whose NVIDIA runtime wheels the owner signed off in `scripts/license_signoff.json` (the release checklist records each sign-off).
+
+## License
+
+Apache-2.0; see [LICENSE](https://github.com/samaust/openftgs/blob/main/LICENSE) and [NOTICE](https://github.com/samaust/openftgs/blob/main/NOTICE).

openftgs-0.1.0/README.md

@@ -0,0 +1,89 @@
+# openftgs
+
+[![CI](https://github.com/samaust/openftgs/actions/workflows/ci.yml/badge.svg)](https://github.com/samaust/openftgs/actions/workflows/ci.yml) [![Coverage](https://raw.githubusercontent.com/samaust/openftgs/main/docs/coverage.svg)](#tests) [![Documentation](https://app.readthedocs.org/projects/openftgs/badge/)](https://openftgs.readthedocs.io)
+
+openftgs is an independent clean-room implementation of FreeTimeGS (Wang et al., CVPR 2025), not affiliated with its authors.
+
+Python package `openftgs`, under the Apache-2.0 license, with the command-line tool `ftgs`.
+
+The documentation is at [openftgs.readthedocs.io](https://openftgs.readthedocs.io): installation, a quickstart, the datasets, the method, the command line and the API reference.
+
+**Status: alpha, version 0.1.0.** The first release, at the end of the plan's milestones M0 to M18: the library and its command line are complete, the differential harness checks them, and the method is validated on openftgs-synth, the owner's CC0-1.0 rendered dataset, rather than on the paper's datasets, so no result is compared with the paper's tables. The [changelog](https://github.com/samaust/openftgs/blob/main/docs/changelog.md) lists what the release contains and its known limitations.
+
+## Clean-room process
+
+The code is written from the paper (arXiv 2506.05348 v2) and from the owner's specification, which was itself written from the paper alone. No code, notebook, checkpoint or issue thread from any other implementation of the method, including the authors' own, is consulted. Third-party software is read only within its license: source code only under permissive licenses that allow commercial use, documentation only for everything else. Every source consulted is logged. The renderer the paper's project page links was read for a documentation-only survey ([#352](https://github.com/samaust/openftgs/issues/352)), and nothing from it is in the package.
+
+- [`docs/spec/openftgs-spec.md`](https://github.com/samaust/openftgs/blob/main/docs/spec/openftgs-spec.md): the specification, the source of truth for the implementation.
+- [`docs/clean-room/source-log.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/source-log.md): every source consulted during the implementation and what was taken from it.
+- [`docs/clean-room/questions.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/questions.md): gaps found in the spec and the provisional decisions made for them. GitHub doesn't render a Markdown file of this size, so it may show only the source; the [raw file](https://github.com/samaust/openftgs/raw/main/docs/clean-room/questions.md) opens as plain text.
+- [`docs/clean-room/milestones.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/milestones.md): the order in which the package was built.
+- [`docs/clean-room/release-checklist.md`](https://github.com/samaust/openftgs/blob/main/docs/clean-room/release-checklist.md): the release-time checks of this version: the license scan, the libraries bundled in dependency wheels, the vendored code's license files, the pretrained weights' terms, the absence of dataset material, and the source-log audit.
+
+## Installation
+
+```bash
+pip install openftgs
+```
+
+openftgs needs Python 3.11 or later. Training needs a CUDA build of PyTorch. On Linux x86_64 the default `torch` wheel on PyPI already is one (torch 2.14.1 bundles the CUDA 13.0 libraries), so the command above installs it. For another CUDA version, or on Windows, install torch from the [PyTorch index](https://pytorch.org/get-started/locally/) before installing openftgs. A CPU build is enough for the tests on the reference backend; CI uses one because its runners have no GPU.
+
+Optional extras:
+
+- `[init]`: the dependencies of the RoMa-based 4D initialization (torchvision, einops and loguru, and tqdm for the code of XFeat that Tiny RoMa loads, Q-276), and of the four DINO perceptual networks (`dinov2_s14`, `dinov2_b14`, `dino_s16` and `dino_b16` as `loss.perc_net`), which run on the vendored RoMa's DINOv2 layers: without the extra, building one is an error that names it (Q-248). ꟻLIP, the default perceptual network, HaarPSI and the two Caffe trunks (`vgg16_oxford`, `alexnet_bvlc`) need no extra. RoMa's optional speed-up, [fused-local-corr](https://github.com/Parskatt/fused-local-corr) (MIT), isn't part of it: every release since 0.3.27 requires one exact torch release (0.3.211 requires torch 2.11.0), so installing one would replace your torch, and the newest unpinned wheel, 0.2.4, built for CUDA 12, doesn't load with torch 2.14.1 (D-85). On Linux, build it from source against the torch you have; this needs the CUDA toolkit (`nvcc`) and a CUDA build of PyTorch already installed:
+
+  ```bash
+  pip install --no-build-isolation git+https://github.com/Parskatt/fused-local-corr@689f2f7303ca0d97ac1f3d1364c732c6d8e889a8
+  ```
+
+  Without it, RoMa runs with its `use_custom_corr` option off and computes the local correlation in plain PyTorch, which is slower.
+- `[fast]`: currently empty. The faster SSIM loss uses [fused-ssim](https://github.com/rahul-goel/fused-ssim) (MIT), which is not on PyPI. Install it from its repository, at the commit openftgs is tested with; this needs the CUDA toolkit (`nvcc`) and a CUDA build of PyTorch already installed:
+
+  ```bash
+  pip install --no-build-isolation git+https://github.com/rahul-goel/fused-ssim@a7c48d6dd7ac6dc39a7958c7c4452e0b10418f38
+  ```
+
+  Without it, the SSIM loss runs in plain PyTorch and gives the same values.
+
+For development, install from a clone of the repository, with the test and lint tools:
+
+```bash
+pip install -e .[dev]
+```
+
+The differential harness is a separate package in `harness/` (`pip install -e harness`), which isn't published.
+
+## Command line
+
+```bash
+ftgs convert n3dv SRC OUT                        # a dataset into a scene manifest (needs ffmpeg)
+ftgs cache OUT/scene.json                        # decode the training frames once (optional)
+ftgs init roma OUT/scene.json init.safetensors   # initial points with RoMa ([init], a CUDA device)
+ftgs train OUT/scene.json --init init.safetensors --out RUN_DIR [--config CFG.json] [--seed S]
+ftgs train OUT/scene.json --init init.safetensors --out RUN_DIR --resume RUN_DIR/checkpoint
+ftgs export RUN_DIR/checkpoint --format ftgs-if --out model.safetensors
+ftgs render model.safetensors --request request.json --out renders.safetensors
+ftgs eval RUN_DIR/checkpoint OUT/scene.json --split test --out METRICS.json
+```
+
+`ftgs <command> --help` lists every option. A run directory holds `config.json`, `train_log.jsonl`, `checkpoint/` and, when the scene has a test split, `METRICS.json`; Ctrl-C saves the checkpoint before stopping. Exit code 0 is success, 2 an invalid input or environment, 3 a numerical failure. See [`docs/modules/cli/command-line.md`](https://github.com/samaust/openftgs/blob/main/docs/modules/cli/command-line.md).
+
+## Tests
+
+```bash
+pytest                       # CPU tests; GPU, dataset and weights tests skip when their resource is missing
+pytest -n auto --dist loadgroup -m "not gpu"      # the CPU tests on every physical core (Q-214)
+pytest -n auto --dist loadgroup -m "not gpu" --cov --cov-report=html   # with coverage, in htmlcov/ (Q-216)
+pytest -m gpu                # tests that need a CUDA device
+OPENFTGS_DATASETS=/path pytest -m dataset         # converter checks on real datasets
+OPENFTGS_ALLOW_DOWNLOADS=1 pytest -m weights      # tests that download pretrained weights
+python scripts/license_scan.py                    # GPL, AGPL, non-commercial fail; others need a sign-off in scripts/license_signoff.json
+```
+
+The coverage badge is the CPU and GPU suites' total on a machine with a CUDA device, the dataset and weights tests left out. `python scripts/coverage_local.py` runs both suites, writes `htmlcov/` and `coverage.xml`, and rewrites `docs/coverage.svg` when both pass; commit the SVG to update the badge (Q-217). CI's `test` job runs the CPU suite on Python 3.14 and on 3.11, the oldest Python openftgs supports (Q-1), and reports its coverage in its summary (Q-216).
+
+The license scan passes in CI's environment (CPU torch) and, for this release, in an environment with CUDA torch, whose NVIDIA runtime wheels the owner signed off in `scripts/license_signoff.json` (the release checklist records each sign-off).
+
+## License
+
+Apache-2.0; see [LICENSE](https://github.com/samaust/openftgs/blob/main/LICENSE) and [NOTICE](https://github.com/samaust/openftgs/blob/main/NOTICE).