Deploy docs with Material Sphinx Theme (#292)

* Deploy docs with Material Sphinx Theme

* Refactor index

* Update requirements for docs

* Update docs configuration

* Fix docs index

* Fixing navigation titles

* Fixing navigation titles

* Fixing configuration

* Minor fixes

* Fixed navigation titles

docs/requirements.txt

@@ -1,5 +1,6 @@
 numpy
 jupyterlab
-sphinx==3.5.4
-insipid-sphinx-theme
-nbsphinx==0.8.7
+ipython
+sphinx
+sphinx-material
+nbsphinx

docs/source/conf.py

@@ -12,16 +12,16 @@
 #
 import os
 import sys
+from datetime import datetime
 
 sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
 
 
 # -- Project information -----------------------------------------------------
 
 project = "yolort"
-copyright = "2020-2021, yolort community"
-author = "Zhiqiang Wang"
-
+copyright = f"{datetime.now().year}, yolort team"
+author = "Zhiqiang Wang, Shiquan Yu, Fidan Kharrasov"
 
 # -- General configuration ---------------------------------------------------
 
@@ -30,8 +30,13 @@
 # ones.
 extensions = [
     "sphinx.ext.autodoc",
+    "sphinx.ext.mathjax",
     "sphinx.ext.napoleon",
+    "sphinx.ext.viewcode",
+    "sphinx.ext.intersphinx",
+    "sphinx.ext.extlinks",
     "nbsphinx",
+    "IPython.sphinxext.ipython_console_highlighting",
 ]
 
 # Add any paths that contain templates here, relative to this directory.
@@ -91,16 +96,45 @@ def setup(app):
     </style>
 """
 
-html_theme = "insipid"
+html_theme = "sphinx_material"
 
-html_context = {
-    "display_github": True,
-    "github_user": "zhiqwang",
-    "github_repo": "yolov5-rt-stack",
-}
+# Material theme options (see theme.conf for more information)
 html_theme_options = {
-    "left_buttons": [],
-    "right_buttons": [
-        "repo-button.html",
-    ],
+    # Set the name of the project to appear in the navigation.
+    "nav_title": "YOLOv5 Runtime Stack",
+    # Set you GA account ID to enable tracking
+    # "google_analytics_account": "UA-XXXXX",
+    # Specify a base_url used to generate sitemap.xml. If not
+    # specified, then no sitemap will be built.
+    "base_url": "https://zhiqwang.com/yolov5-rt-stack",
+    # Set the color and the accent color
+    "color_primary": "blue",
+    "color_accent": "light-blue",
+    # Set the repo location to get a badge with stats
+    "repo_url": "https://github.com/zhiqwang/yolov5-rt-stack/",
+    "repo_name": "yolort",
+    # Visible levels of the global TOC; -1 means unlimited
+    "globaltoc_depth": 3,
+    # If False, expand all TOC entries
+    "globaltoc_collapse": False,
+    # If True, show hidden TOC entries
+    "globaltoc_includehidden": False,
+    # Text to appear at the top of the home page in a "hero" div.
+    "heroes": {
+        # We can have heroes for the home pages of training and inferencing in future.
+        "index": "A runtime stack for object detection on specialized accelerators."
+    },
 }
+
+# Disable show source link.
+html_show_sourcelink = False
+
+# Custom sidebar templates, must be a dictionary that maps document names
+# to template names.
+#
+# The default sidebars (for documents that don't match any pattern) are
+# defined by theme itself.  Builtin themes are using these templates by
+# default: ``['localtoc.html', 'relations.html', 'sourcelink.html',
+# 'searchbox.html']``.
+#
+html_sidebars = {"**": ["logo-text.html", "globaltoc.html", "localtoc.html", "searchbox.html"]}

docs/source/index.rst

@@ -1,25 +1,18 @@
 Welcome to yolort's documentation!
 ==================================
 
-Runtime stack for YOLOv5 on specialized accelerators such as ``libtorch``,
-``onnxruntime``, ``tvm`` and ``ncnn``.
-
-.. image:: _static/yolort_logo.png
-   :width: 400px
-   :align: center
-
-
 .. _what-is-yolort:
 
 **What is yolort?**
 
 ``yolort`` focus on making the training and inference of the object detection task
 integrate more seamlessly together. ``yolort`` now adopts the same model
-structure as the official YOLOv5. The significant difference is that we adopt
+structure as the official ``YOLOv5``. The significant difference is that we adopt
 the dynamic shape mechanism, and within this, we can embed both pre-processing
 (``letterbox``) and post-processing (``nms``) into the model graph, which
 simplifies the deployment strategy. In this sense, ``yolort`` makes it possible
-to be deployed more friendly on ``LibTorch``, ``ONNXRuntime``, ``TVM`` and so on.
+to be deployed more friendly on ``LibTorch``, ``ONNX Runtime``, ``TensorRT``, ``TVM``
+and so on.
 
 .. _about-the-code:
 
@@ -33,11 +26,11 @@ Follow the design principle of `detr <https://github.com/facebookresearch/detr>`
    not require complex libraries for training and inference.
 
 ``yolort`` is very simple to implement and experiment with. You like the implementation
-of torchvision's faster-rcnn, retinanet or detr? You like yolov5? You love ``yolort``!
+of torchvision's faster-rcnn, retinanet or detr? You like yolov5? You love yolort!
 
 .. _quick-get-stated:
 
-**Quick Get started**
+**Introduction to yolort**
 
 Read a source of image(s) and detect its objects:
 
@@ -54,20 +47,49 @@ Read a source of image(s) and detect its objects:
    # Perform inference on a list of image files
    predictions = model.predict(["bus.jpg", "zidane.jpg"])
 
+**Loading checkpoint from official yolov5**
+
+And we support loading the trained weights from YOLOv5:
+
+.. code:: python
+
+   from yolort.models import YOLOv5
+   from yolort.v5 import attempt_download
+
+   # will downloaded from 'https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5n6.pt'
+   model_path = "yolov5n6.pt"
+   checkpoint_path = attempt_download(model_path)
+
+   model = YOLOv5.load_from_yolov5(checkpoint_path, score_thresh=0.25)
+
+   model.eval()
+   img_path = "bus.jpg"
+   predictions = model.predict(img_path)
+
+**Use Cases and Solutions**
+
 .. toctree::
    :maxdepth: 2
-   :caption: Get started
+   :titlesonly:
+   :caption: Getting Started
 
    installation
 
 .. toctree::
    :maxdepth: 2
+   :titlesonly:
    :caption: Tutorials
 
    notebooks/inference-pytorch-export-libtorch
    notebooks/how-to-align-with-ultralytics-yolov5
    notebooks/anchor-label-assignment-visualization
    notebooks/model-graph-visualization
+
+.. toctree::
+   :maxdepth: 2
+   :titlesonly:
+   :caption: Deployment
+
    notebooks/export-onnx-inference-onnxruntime
    notebooks/onnx-graphsurgeon-inference-tensorrt
    notebooks/export-relay-inference-tvm

docs/source/installation.rst

@@ -1,4 +1,4 @@
-Installation
+Install yolort
 ============
 
 .. _required:

notebooks/export-onnx-inference-onnxruntime.ipynb

@@ -4,7 +4,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Deploying yolort on ONNXRuntime"
+    "# Deploying yolort on ONNX Runtime"
    ]
   },
   {