docs: add per-project Citation pages, SEO metadata, and content refinements

- Add a Citation page to each project (DeepLens, End2endImaging, DiffTMM) as
  the last left-menu item; move DiffTMM's citation off the Overview.
- SEO: inject per-page Open Graph + Twitter Card meta and a WebSite JSON-LD
  block (overrides/main.html), add unique per-page meta descriptions on the
  landing pages, a robots.txt referencing the sitemap, and theme.language.
- End2endImaging Overview: add Key Features and Advanced Features sections;
  drop the redundant Architecture heading and keep the pipeline diagram.
- DeepLens Overview: reword the use cases (differentiable optics, automated
  optical design, computational imaging) and relabel the diagram input as a
  light source.
- DiffTMM Overview: add an input/output diagram, placed after the use cases.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: singer-yang

docs/deeplens/citation.md

@@ -0,0 +1,69 @@
+# Citation
+
+If you use **DeepLens** in your research, please cite the relevant papers below.
+
+**Differentiable ray tracing** was developed by the [KAUST Computational Imaging Group](https://vccimaging.org/). The first version of the differentiable ray tracer was implemented by [Dr. Congli Wang](https://congliwang.github.io/) on [Mitsuba2](https://github.com/mitsuba-renderer/mitsuba2):
+
+```bibtex
+@article{sun2021end,
+  title={End-to-end complex lens design with differentiable ray tracing},
+  author={Sun, Qilin and Wang, Congli and Qiang, Fu and Xiong, Dun and Wolfgang, Heidrich},
+  journal={ACM Trans. Graph},
+  volume={40},
+  number={4},
+  pages={1--13},
+  year={2021}
+}
+```
+
+Later, Dr. Congli Wang implemented the first PyTorch version of the differentiable ray tracer ([**dO**](https://github.com/vccimaging/DiffOptics)):
+
+```bibtex
+@article{wang2022differentiable,
+  title={do: A differentiable engine for deep lens design of computational imaging systems},
+  author={Wang, Congli and Chen, Ni and Heidrich, Wolfgang},
+  journal={IEEE Transactions on Computational Imaging},
+  volume={8},
+  pages={905--916},
+  year={2022},
+  publisher={IEEE}
+}
+```
+
+Currently, [Xinge Yang](https://singer-yang.github.io/) builds and maintains [**DeepLens**](https://github.com/AI4Optics/DeepLens). The automated lens design work demonstrates the capability of differentiable optics compared to classical optical design:
+
+```bibtex
+@article{yang2024curriculum,
+  title={Curriculum learning for ab initio deep learned refractive optics},
+  author={Yang, Xinge and Fu, Qiang and Heidrich, Wolfgang},
+  journal={Nature communications},
+  volume={15},
+  number={1},
+  pages={6572},
+  year={2024},
+  publisher={Nature Publishing Group UK London}
+}
+```
+
+A differentiable **ray-wave model** is introduced for simulating and optimizing hybrid refractive-diffractive lenses:
+
+```bibtex
+@inproceedings{yang2024end,
+  title={End-to-end hybrid refractive-diffractive lens design with differentiable ray-wave model},
+  author={Yang, Xinge and Souza, Matheus and Wang, Kunyi and Chakravarthula, Praneeth and Fu, Qiang and Heidrich, Wolfgang},
+  booktitle={SIGGRAPH Asia 2024 Conference Papers},
+  pages={1--11},
+  year={2024}
+}
+```
+
+Differentiable **non-sequential** ray tracing and **polarization tracing** are developed in:
+
+```bibtex
+@article{yang2026waveguide,
+  title={End-to-end differentiable design of geometric waveguide displays},
+  author={Yang, Xinge and Liu, Zhaocheng and Nie, Zhaoyu and Fan, Qingyuan and Shi, Zhimin and Bonar, Jim and Heidrich, Wolfgang},
+  journal={arXiv preprint arXiv:2601.04370},
+  year={2026}
+}
+```

docs/deeplens/index.md

@@ -1,11 +1,21 @@
+---
+description: DeepLens is a PyTorch differentiable optical lens simulator — geometric ray tracing, diffractive wave optics, PSF computation, automated lens design, and end-to-end computational imaging.
+---
+
 # DeepLens
 
-DeepLens is a PyTorch-based differentiable optical lens simulator for end-to-end computational imaging, supporting multiple optical models — geometric ray tracing, diffractive wave propagation, hybrid ray-wave, and surrogate PSF networks. It can be used for end-to-end optics-algorithm co-design, gradient-based automated optical design, and synthetic dataset generation via image simulation, letting researchers rapidly prototype and optimize custom optical systems.
+DeepLens is a PyTorch-based differentiable optical lens simulator for end-to-end computational imaging, supporting multiple optical models — geometric ray tracing, diffractive wave propagation, hybrid ray-wave, and surrogate PSF networks.
+
+DeepLens supports three main use cases:
+
+- **Differentiable optics** — accurate and efficient gradient computation for optical parameters via differentiable simulation and backpropagation.
+- **Automated optical design** — fully automated optical design driven by gradient information and advanced optimization algorithms.
+- **Computational imaging** — physically accurate image simulation and end-to-end optimization with downstream image reconstruction algorithms.
 
 DeepLens computes the point spread function (PSF) of an optical lens in a fully differentiable manner:
 
 ```
-Point source → [ DeepLens ] → PSF
+Light source → [ DeepLens ] → PSF
 ```
 
 DeepLens also serves as the differentiable optics engine in an end-to-end computational imaging pipeline such as [End2endImaging](https://github.com/vccimaging/End2endImaging), where the optics, sensor, and a reconstruction network form a single differentiable graph that can be optimized jointly:

docs/difftmm/citation.md

@@ -0,0 +1,18 @@
+# Citation
+
+If you use DiffTMM in your research, please cite the accompanying paper:
+
+> X. Yang, Z. Liu, Z. Nie, Q. Fan, Z. Shi, J. Bonar, and W. Heidrich,
+> "End-to-end differentiable design of geometric waveguide displays,"
+> *arXiv preprint* [arXiv:2601.04370](https://arxiv.org/abs/2601.04370) (2026).
+
+```bibtex
+@article{yang2026waveguide,
+  title={End-to-end differentiable design of geometric waveguide displays},
+  author={Yang, Xinge and Liu, Zheng and Nie, Zhanghao and Fan, Qiang and Shi, Zheng and Bonar, James and Heidrich, Wolfgang},
+  journal={arXiv preprint arXiv:2601.04370},
+  year={2026}
+}
+```
+
+DiffTMM is released under the [Apache License 2.0](https://github.com/AI4Optics/DiffTMM/blob/main/LICENSE).

docs/difftmm/index.md

@@ -1,3 +1,7 @@
+---
+description: DiffTMM is a differentiable transfer matrix method (TMM) in PyTorch for multi-layer thin-film optics — GPU-batched forward simulation and gradient-based inverse design of optical coatings.
+---
+
 # DiffTMM
 
 DiffTMM is a differentiable [Transfer Matrix Method](https://en.wikipedia.org/wiki/Transfer-matrix_method_(optics))
@@ -13,6 +17,12 @@ DiffTMM supports two main use cases:
 - **Inverse design** — optimize layer thicknesses (and stack geometry) end-to-end
   with PyTorch autograd to match measured or target optical properties.
 
+DiffTMM computes the Fresnel coefficients of a multi-layer thin-film stack in a fully differentiable manner:
+
+```
+Incident angle, wavelength → [ DiffTMM ] → Fresnel coefficients (ts, tp, rs, rp)
+```
+
 ## Why DiffTMM
 
 The transfer matrix method is the standard tool for multi-layer thin-film optics,
@@ -47,6 +57,8 @@ Refractive indices can be plain numbers, complex values (with loss), or real
 [material names](api/material.md) (`"N-BK7"`, `"SiO2"`, `"Ag"`, …) resolved to
 wavelength-dependent dispersion from bundled catalogs.
 
+See the [Citation](citation.md) page for how to cite DiffTMM.
+
 ## Getting Started
 
 <div class="grid cards" markdown>
@@ -92,13 +104,3 @@ wavelength-dependent dispersion from bundled catalogs.
     [:octicons-arrow-right-24: See examples](examples/index.md)
 
 </div>
-
-## Citation
-
-If you use DiffTMM in your research, please cite the accompanying paper:
-
-> X. Yang, Z. Liu, Z. Nie, Q. Fan, Z. Shi, J. Bonar, and W. Heidrich,
-> "End-to-end differentiable design of geometric waveguide displays,"
-> *arXiv preprint* [arXiv:2601.04370](https://arxiv.org/abs/2601.04370) (2026).
-
-DiffTMM is released under the [Apache License 2.0](https://github.com/AI4Optics/DiffTMM/blob/main/LICENSE).

docs/end2endimaging/citation.md

@@ -0,0 +1,20 @@
+# Citation
+
+If you use **End2endImaging** in your research, please cite:
+
+> X. Yang, Q. Fu, and W. Heidrich, "Curriculum learning for ab initio deep
+> learned refractive optics," *Nature Communications* **15**(1), 6572 (2024).
+> [doi:10.1038/s41467-024-50835-7](https://www.nature.com/articles/s41467-024-50835-7)
+
+```bibtex
+@article{yang2024curriculum,
+  title={Curriculum learning for ab initio deep learned refractive optics},
+  author={Yang, Xinge and Fu, Qiang and Heidrich, Wolfgang},
+  journal={Nature communications},
+  volume={15},
+  number={1},
+  pages={6572},
+  year={2024},
+  publisher={Nature Publishing Group UK London}
+}
+```

docs/end2endimaging/index.md

@@ -1,3 +1,7 @@
+---
+description: End2endImaging is a PyTorch framework for end-to-end differentiable computational imaging — joint co-design of optics, sensor, ISP, and neural image reconstruction.
+---
+
 # End2endImaging
 
 End2endImaging models the full imaging pipeline — optics, sensor, and image processing — as a differentiable computation graph built on PyTorch. This enables gradient-based optimization of camera systems from lens surfaces all the way through neural image reconstruction.
@@ -7,8 +11,6 @@ End2end Imaging targets two main applications:
 - **High-fidelity image simulation** — physically accurate rendering of camera captures for synthetic dataset generation and physical AI.
 - **End-to-end optics–algorithm co-design** — joint optimization of lens surfaces and reconstruction algorithms for computational imaging.
 
-## Architecture
-
 The imaging pipeline composes three differentiable stages — optics (DeepLens), sensor, and reconstruction network:
 
 ```
@@ -21,7 +23,19 @@ Scene Image → [ DeepLens ] → Spectral Image → [ Sensor ] → Raw Image →
                 PSFNetLens
 ```
 
-See [Architecture](architecture.md) for the full module breakdown.
+## Key Features
+
+- **End-to-End Differentiable Pipeline** — the full camera pipeline (optics, sensor, ISP, and network) as a single differentiable graph. Gradients flow from downstream task losses (reconstruction, detection, classification) back through the network, sensor noise model, and ISP into the optical design parameters, enabling hardware–software co-optimization.
+- **Differentiable Optics** — built on [DeepLens](../deeplens/index.md): `GeoLens`, `HybridLens`, `DiffractiveLens`, `PSFNetLens`, and `DefocusLens` for differentiable ray tracing and wave-optics simulation and design.
+- **Sensor & ISP Simulation** — physically accurate sensor simulation with a Bayer CFA and a read/shot-noise model, plus a composable, fully differentiable ISP pipeline (black level, white balance, demosaicing, color correction, gamma, tone mapping) where every stage is a `torch.nn.Module`.
+- **Neural Networks** — built-in image-reconstruction networks (`NAFNet`, `UNet`, `Restormer`) for restoring clean images from degraded captures, plus PSF surrogate networks (`MLP`, `SIREN`) for fast PSF prediction during training.
+
+## Advanced Features
+
+Additional capabilities, available upon request:
+
+- **GPU Kernel Acceleration** — >10× speedup and >90% GPU memory reduction with custom GPU kernels across NVIDIA and AMD platforms.
+- **Distributed Optimization** — distributed simulation and optimization for billions of rays and high-resolution (>100k) diffractive propagation.
 
 ## Getting Started
 

docs/index.md

@@ -1,4 +1,5 @@
 ---
+description: AI4Optics is a family of open-source, PyTorch-based differentiable optics tools for optical design and computational imaging — DeepLens lens simulation, End2endImaging, and DiffTMM thin-film TMM.
 hide:
   - navigation
   - toc

docs/robots.txt

@@ -0,0 +1,4 @@
+User-agent: *
+Allow: /
+
+Sitemap: https://ai4optics.github.io/AI4Optics-docs/sitemap.xml

mkdocs.yml

@@ -15,6 +15,7 @@ docs_dir: docs
 theme:
   name: material
   custom_dir: overrides
+  language: en
   favicon: assets/logo.png
   icon:
     repo: fontawesome/brands/github
@@ -121,6 +122,7 @@ nav:
               - Pupil Field: deeplens/examples/pupil_field.md
               - Multi-order Diffraction: deeplens/examples/multi_order.md
               - Diffractive Surfaces: deeplens/examples/diffractive_surfaces.md
+      - Citation: deeplens/citation.md
   - End2endImaging:
       - Overview: end2endimaging/index.md
       - Get Started:
@@ -139,6 +141,7 @@ nav:
           - Defocus Deblur: end2endimaging/examples/defocus_deblur.md
           - End-to-End Lens Design: end2endimaging/examples/end2end_lens_design.md
           - Task-Driven Lens Design: end2endimaging/examples/task_driven.md
+      - Citation: end2endimaging/citation.md
   - DiffTMM:
       - Overview: difftmm/index.md
       - Get Started:
@@ -158,4 +161,5 @@ nav:
           - Inverse Design: difftmm/examples/inverse_design.md
           - Real Materials: difftmm/examples/real_materials.md
           - Incoherent Films: difftmm/examples/incoherent_films.md
+      - Citation: difftmm/citation.md
   - Contribute: contribute.md

overrides/main.html

@@ -0,0 +1,39 @@
+{% extends "base.html" %}
+
+{#-
+  SEO additions injected into <head> via Material's `extrahead` block:
+  per-page Open Graph + Twitter Card meta and a WebSite JSON-LD node.
+
+  Descriptions come from each page's front-matter `description:` when present,
+  otherwise the global `site_description`. Title, URL, and image are absolute
+  (built from `site_url`) so social scrapers and crawlers resolve them. The
+  JSON-LD values use the `tojson` filter, which emits HTML-safe JSON strings.
+-#}
+{% block extrahead %}
+  {% set _title = (page.title ~ " - " ~ config.site_name) if (page and page.title and not page.is_homepage) else config.site_name %}
+  {% set _desc = page.meta.description if (page and page.meta and page.meta.description) else config.site_description %}
+  {% set _url = page.canonical_url if (page and page.canonical_url) else config.site_url %}
+  {% set _image = config.site_url ~ "assets/logo.png" %}
+
+  <meta property="og:type" content="website">
+  <meta property="og:site_name" content="{{ config.site_name }}">
+  <meta property="og:title" content="{{ _title }}">
+  <meta property="og:description" content="{{ _desc }}">
+  <meta property="og:url" content="{{ _url }}">
+  <meta property="og:image" content="{{ _image }}">
+
+  <meta name="twitter:card" content="summary">
+  <meta name="twitter:title" content="{{ _title }}">
+  <meta name="twitter:description" content="{{ _desc }}">
+  <meta name="twitter:image" content="{{ _image }}">
+
+  <script type="application/ld+json">
+  {
+    "@context": "https://schema.org",
+    "@type": "WebSite",
+    "name": {{ config.site_name | tojson }},
+    "url": {{ config.site_url | tojson }},
+    "description": {{ config.site_description | tojson }}
+  }
+  </script>
+{% endblock %}