[fbsync] Vectorize box decoding in FCOS (#6203)

Summary:
* basic structure

* added constrains

* fixed errors

* thanks to vadim!

* addressing the comments and added docstrign

* Apply suggestions from code review

Reviewed By: jdsgomes

Differential Revision: D37643906

fbshipit-source-id: 13b15e886d23e259a5ed6926677cac55f160f206

Co-authored-by: Vasilis Vryniotis <[email protected]>

Author: datumbox

torchvision/models/detection/_utils.py

@@ -300,6 +300,42 @@ def decode_single(self, rel_codes: Tensor, boxes: Tensor) -> Tensor:
         pred_boxes = torch.stack((pred_boxes1, pred_boxes2, pred_boxes3, pred_boxes4), dim=1)
         return pred_boxes
 
+    def decode_all(self, rel_codes: Tensor, boxes: List[Tensor]) -> Tensor:
+        """
+        Vectorized version of `decode_single` method.
+
+        Args:
+            rel_codes (Tensor): encoded boxes
+            boxes (List[Tensor]): List of reference boxes.
+
+        Returns:
+            Tensor: the predicted boxes with the encoded relative box offsets.
+
+        .. note::
+            This method assumes that ``rel_codes`` and ``boxes`` have same size for 0th dimension. i.e. ``len(rel_codes) == len(boxes)``.
+
+        """
+
+        boxes = torch.stack(boxes).to(dtype=rel_codes.dtype)
+
+        ctr_x = 0.5 * (boxes[..., 0] + boxes[..., 2])
+        ctr_y = 0.5 * (boxes[..., 1] + boxes[..., 3])
+
+        if self.normalize_by_size:
+            boxes_w = boxes[..., 2] - boxes[..., 0]
+            boxes_h = boxes[..., 3] - boxes[..., 1]
+
+            list_box_size = torch.stack((boxes_w, boxes_h, boxes_w, boxes_h), dim=-1)
+            rel_codes = rel_codes * list_box_size
+
+        pred_boxes1 = ctr_x - rel_codes[..., 0]
+        pred_boxes2 = ctr_y - rel_codes[..., 1]
+        pred_boxes3 = ctr_x + rel_codes[..., 2]
+        pred_boxes4 = ctr_y + rel_codes[..., 3]
+
+        pred_boxes = torch.stack((pred_boxes1, pred_boxes2, pred_boxes3, pred_boxes4), dim=-1)
+        return pred_boxes
+
 
 class Matcher:
     """

torchvision/models/detection/fcos.py

@@ -87,14 +87,12 @@ def compute_loss(
         loss_cls = sigmoid_focal_loss(cls_logits, gt_classes_targets, reduction="sum")
 
         # regression loss: GIoU loss
-        # TODO: vectorize this instead of using a for loop
-        pred_boxes = [
-            self.box_coder.decode_single(bbox_regression_per_image, anchors_per_image)
-            for anchors_per_image, bbox_regression_per_image in zip(anchors, bbox_regression)
-        ]
+
+        pred_boxes = self.box_coder.decode_all(bbox_regression, anchors)
+
         # amp issue: pred_boxes need to convert float
         loss_bbox_reg = generalized_box_iou_loss(
-            torch.stack(pred_boxes)[foregroud_mask].float(),
+            pred_boxes[foregroud_mask],
             torch.stack(all_gt_boxes_targets)[foregroud_mask],
             reduction="sum",
         )