revamp FCOS doc (#5901)

* refactor doc

* refactor code

* add fcos.rst

* add fcos
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

docs/source/models/fcos.rst

+FCOS
+=========
+
+.. currentmodule:: torchvision.models.detection
+
+The RetinaNet model is based on the `FCOS: Fully Convolutional One-Stage Object Detection
+<https://arxiv.org/abs/1904.01355>`__ paper.
+
+Model builders
+--------------
+
+The following model builders can be used to instantiate a FCOS model, with or
+without pre-trained weights. All the model buidlers internally rely on the
+``torchvision.models.detection.fcos.FCOS`` base class. Please refer to the `source code
+<https://github.com/pytorch/vision/blob/main/torchvision/models/detection/fcos.py>`_ for
+more details about this class.
+
+.. autosummary::
+    :toctree: generated/
+    :template: function.rst
+
+    fcos_resnet50_fpn

docs/source/models_new.rst

@@ -93,8 +93,9 @@ weights:
 .. toctree::
    :maxdepth: 1
 
-   models/retinanet
+   models/fcos
    models/mask_rcnn
+   models/retinanet
 
 Table of all available detection weights
 ----------------------------------------

torchvision/models/detection/fcos.py

@@ -679,7 +679,7 @@ def fcos_resnet50_fpn(
     """
     Constructs a FCOS model with a ResNet-50-FPN backbone.
 
-    Reference: `"FCOS: Fully Convolutional One-Stage Object Detection" <https://arxiv.org/abs/1904.01355>`_.
+    Reference: `FCOS: Fully Convolutional One-Stage Object Detection <https://arxiv.org/abs/1904.01355>`_.
 
     The input to the model is expected to be a list of tensors, each of shape ``[C, H, W]``, one for each
     image, and should be in ``0-1`` range. Different images can have different sizes.
@@ -715,13 +715,25 @@ def fcos_resnet50_fpn(
         >>> predictions = model(x)
 
     Args:
-        weights (FCOS_ResNet50_FPN_Weights, optional): The pretrained weights for the model
+        weights (:class:`~torchvision.models.detection.FCOS_ResNet50_FPN_Weights`, optional): The
+            pretrained weights to use. See
+            :class:`~torchvision.models.detection.FCOS_ResNet50_FPN_Weights`
+            below for more details, and possible values. By default, no
+            pre-trained weights are used.
         progress (bool): If True, displays a progress bar of the download to stderr
         num_classes (int, optional): number of output classes of the model (including the background)
-        weights_backbone (ResNet50_Weights, optional): The pretrained weights for the backbone
+        weights_backbone (:class:`~torchvision.models.ResNet50_Weights`, optional): The pretrained weights for
+            the backbone.
         trainable_backbone_layers (int, optional): number of trainable (not frozen) resnet layers starting
             from final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are
             trainable. If ``None`` is passed (the default) this value is set to 3. Default: None
+        **kwargs: parameters passed to the ``torchvision.models.detection.FCOS``
+            base class. Please refer to the `source code
+            <https://github.com/pytorch/vision/blob/main/torchvision/models/detection/fcos.py>`_
+            for more details about this class.
+
+    .. autoclass:: torchvision.models.detection.FCOS_ResNet50_FPN_Weights
+        :members:
     """
     weights = FCOS_ResNet50_FPN_Weights.verify(weights)
     weights_backbone = ResNet50_Weights.verify(weights_backbone)