Problem about calculation of loss

[Retiina]

Dear author, I am really puzzled about the loss calculation in the following code segmentation in free_anchor_loss.py, could you explain to me please?

Thanks in advance!

with torch.set_grad_enabled(False):
    box_localization = self.box_coder.decode(box_regression_, anchors_.bbox)
    object_box_iou = boxlist_iou(
        targets_,
        BoxList(box_localization, anchors_.size, mode='xyxy')
    )
    H = object_box_iou.max(dim=1, keepdim=True).values.clamp(
        min=self.bbox_threshold + 1e-12)
    object_box_prob = (
            (object_box_iou - self.bbox_threshold) / (H - self.bbox_threshold)
    ).clamp(min=0, max=1)

    indices = torch.stack(
        [torch.arange(len(labels_)).type_as(labels_), labels_], dim=0)

    """
    to implement image_box_iou = torch.sparse.max(
                      torch.sparse_coo_tensor(indices, object_box_iou), dim=0
                 )
    """
    # start
    indices = torch.nonzero(torch.sparse.sum(
        torch.sparse_coo_tensor(indices, object_box_prob), dim=0
    ).to_dense()).t_()

    if indices.numel() == 0:
        image_box_prob = torch.zeros(anchors_.bbox.size(0),
                                     self.num_classes).type_as(object_box_prob)
    else:
        nonzero_box_prob = torch.where(
            (labels_.unsqueeze(dim=-1) == indices[0]),
            object_box_prob[:, indices[1]],
            torch.tensor([0]).type_as(object_box_prob)
        ).max(dim=0).values

        image_box_prob = torch.sparse_coo_tensor(
            indices.flip([0]), nonzero_box_prob,
            size=(anchors_.bbox.size(0), self.num_classes)
        ).to_dense()
    # end

[zhangxiaosong18]

This part of the code is really complicated, I have added more explanations and correspondences with formulas, hope it can help you.

with torch.set_grad_enabled(False):
    # box_localization: a_{j}^{loc}, shape: [j, 4]
    box_localization = self.box_coder.decode(box_regression_, anchors_.bbox)

    # object_box_iou: IoU_{ij}^{loc}, shape: [i, j]
    object_box_iou = boxlist_iou(
        targets_,
        BoxList(box_localization, anchors_.size, mode='xyxy')
    )

    t1 = self.bbox_threshold
    t2 = object_box_iou.max(dim=1, keepdim=True).values.clamp(min=t1 + 1e-12)

    # object_box_prob: P{a_{j} -> b_{i}}, shape: [i, j]
    object_box_prob = (
        (object_box_iou - t1) / (t2 - t1)
    ).clamp(min=0, max=1)

    indices = torch.stack([torch.arange(len(labels_)).type_as(labels_), labels_], dim=0)

    # object_cls_box_prob: P{a_{j} -> b_{i}}, shape: [i, c, j]
    object_cls_box_prob = torch.sparse_coo_tensor(indices, object_box_prob)

    # image_box_iou: P{a_{j} \in A_{+}}, shape: [j, c]
    """
    from "start" to "end" implement:
    
    image_box_iou = torch.sparse.max(object_cls_box_prob, dim=0).t()
    
    """
    # start
    indices = torch.nonzero(torch.sparse.sum(
        object_cls_box_prob, dim=0
    ).to_dense()).t_()

    if indices.numel() == 0:
        image_box_prob = torch.zeros(anchors_.bbox.size(0), self.num_classes).type_as(object_box_prob)
    else:
        nonzero_box_prob = torch.where(
            (labels_.unsqueeze(dim=-1) == indices[0]),
            object_box_prob[:, indices[1]],
            torch.tensor([0]).type_as(object_box_prob)
        ).max(dim=0).values

        image_box_prob = torch.sparse_coo_tensor(
            indices.flip([0]), nonzero_box_prob,
            size=(anchors_.bbox.size(0), self.num_classes)
        ).to_dense()
    # end

[Retiina]

Thank you again sir!