DreamLens
Copyright 2026 DreamLens Contributors

DreamLens's original source code is licensed under the Apache License, Version
2.0, as provided in LICENSE. This NOTICE records research provenance and
third-party materials; it does not replace or expand any third-party license.

Research foundations
--------------------

1. Chris Olah, Alexander Mordvintsev, and Ludwig Schubert, "Feature
   Visualization," Distill, 2017.
   https://distill.pub/2017/feature-visualization/

2. Thomas Fel, Thibaut Boissin, Victor Boutin, Agustin Picard, Paul Novello,
   Julien Colin, Drew Linsley, Tom Rousseau, Remi Cadene, Lore Goetschalckx,
   Laurent Gardes, and Thomas Serre, "Unlocking Feature Visualization for Deep
   Network with MAgnitude Constrained Optimization," NeurIPS, 2023.
   https://proceedings.neurips.cc/paper_files/paper/2023/hash/76d2f8e328e1081c22a77ca0fa330ca5-Abstract-Conference.html

3. Chris Hamblin, Thomas Fel, Srijani Saha, Talia Konkle, and George Alvarez,
   "Feature Accentuation: Revealing 'What' Features Respond to in Natural
   Images," arXiv:2402.10039, 2024.
   https://arxiv.org/abs/2402.10039

DreamLens is an independent educational and research implementation. It is not
an official implementation of, endorsed by, or affiliated with the authors,
their institutions, Distill, NeurIPS, or arXiv. Research concepts are cited;
paper text and figures are not redistributed by the Python package.

Software acknowledgements
-------------------------

- Lucent (Apache-2.0): https://github.com/greentfrapp/lucent
- Xplique (Apache-2.0): https://github.com/deel-ai/xplique
- Faccent (MIT, as stated by its repository):
  https://github.com/chrishamblin7/faccent
- PyTorch and torchvision: https://pytorch.org/

DreamLens does not import Lucent, Xplique, or Faccent at runtime. Their names
identify research/reference software and do not imply endorsement.

Data and example assets
-----------------------

- src/dreamlens/data/clean_decorrelated.npy is the natural-image Fourier
  magnitude used by DreamLens MaCo and optional phase-based feature
  accentuation. It was obtained from:
  https://storage.googleapis.com/serrelab/loupe/spectrums/imagenet_decorrelated.npy
  SHA-256: a4810ea049ef9a0fe4e3f26660188e53222281879b333e8fd61377f7491aafc8
  This data asset is not relicensed by the DreamLens Apache-2.0 license.

- learning_inputs/iguana.jpg and learning_inputs/fox.jpg are repository-only
  tutorial inputs sourced from the public Faccent example directory. The
  source repository does not separately identify their photographer or image
  license. They are excluded from the PyPI distributions and are not covered
  by the DreamLens Apache-2.0 license. Source:
  https://github.com/chrishamblin7/faccent/tree/main/test_images
  iguana.jpg SHA-256:
  976980e1e49b15a1a87f9ff22c40ca66f8166bcd1b09807b5220af6f2df1602a
  fox.jpg SHA-256:
  cc908e850f84f7363e67712d697f8d10a6ea94db6e7b9ff61d1eccf9de92a01e

Third-party model weights, datasets, images, names, and publications remain
subject to their respective owners' terms. Educational or research intent does
not by itself waive copyright, trademark, patent, privacy, or license
obligations. Users and redistributors are responsible for confirming that
their use complies with applicable terms and law.
