loss方法

0、方法的输入与输出

输入

cls_scores(list[Tensor])： 每个预测框的类别得分。每个tensor的形状为：(N, num_anchors * num_classes, H,
W)，其中N表示batch_size，num_anchors表示基础anchor的数量，mun_classes表示类别数量，H、W分别表示特征图的高和宽。
bbox_preds (list[Tensor]): 表示RPN输出预测框的位置，根据该值可以计算出预测框在原图上的位置。每个tensor的形状为： (N, num_anchors * 4,H, W)
gt_bboxes (list[Tensor]): 表示真实标注框的位置。形状为 (num_gts, 4)的tensor，每一行表示标注框的左上角点和右下角点。
gt_labels (list[Tensor]): 表示每个标注框的类别。
img_metas (list[dict]): 表示输入图片的相关信息。
gt_bboxes_ignore (None | list[Tensor]): 表示被忽略的标注，默认为None。

输出

dict[str, Tensor]: 一个loss字典。其中包括类别损失和边界框回归损失。

1、根据输入特征的尺度与图片的相关信息在相应的设别上生成anchor

        featmap_sizes = [featmap.size()[-2:] for featmap in cls_scores] 
        assert len(featmap_sizes) == self.anchor_generator.num_levels
        device = cls_scores[0].device

        anchor_list, valid_flag_list = self.get_anchors(featmap_sizes, img_metas, device=device)

获取输入特征的尺寸。
检查输入数据是否正确。
获取当前工作设备。
调用get_anchors()方法获取对应的anchor列表。

这里的anchor_list就是生成的anchor列表，valid_flag_list表示各个anchor是否合法。

2、根据真实标签构造一个tensor

        label_channels = self.cls_out_channels if self.use_sigmoid_cls else 1
        cls_reg_targets = self.get_targets(
            anchor_list,
            valid_flag_list,
            gt_bboxes,
            img_metas,
            gt_bboxes_ignore_list=gt_bboxes_ignore,
            gt_labels_list=gt_labels,
            label_channels=label_channels)
        if cls_reg_targets is None:
            return None
        (labels_list, label_weights_list, bbox_targets_list, bbox_weights_list, num_total_pos, num_total_neg) = cls_reg_targets
        num_total_samples = (num_total_pos + num_total_neg if self.sampling else num_total_pos)

两个维度相同的tensor才能计算损失。该部分待完善。

3、整理生成anchor的格式，使之与构造的tensor相匹配

        # anchor number of multi levels
        num_level_anchors = [anchors.size(0) for anchors in anchor_list[0]]
        # concat all level anchors and flags to a single tensor
        concat_anchor_list = []
        for i in range(len(anchor_list)):
            concat_anchor_list.append(torch.cat(anchor_list[i]))
        all_anchor_list = images_to_levels(concat_anchor_list, num_level_anchors)

4、计算预测值与真实标签之间的距离

        losses_cls, losses_bbox = multi_apply(
            self.loss_single,
            cls_scores,
            bbox_preds,
            all_anchor_list,
            labels_list,
            label_weights_list,
            bbox_targets_list,
            bbox_weights_list,
            num_total_samples=num_total_samples)

loss_single方法

0、输入与输出

输入

cls_score (Tensor): Box scores for each scale level has shape (N, num_anchors * num_classes, H, W).
bbox_pred (Tensor): Box energies / deltas for each scale level with shape (N, num_anchors * 4, H, W).
anchors (Tensor): Box reference for each scale level with shape (N, num_total_anchors, 4).
labels (Tensor): Labels of each anchors with shape (N, num_total_anchors).
label_weights (Tensor): Label weights of each anchor with shape (N, num_total_anchors)
bbox_targets (Tensor): BBox regression targets of each anchor wight shape (N, num_total_anchors, 4).
bbox_weights (Tensor): BBox regression loss weights of each anchor with shape (N, num_total_anchors, 4).
num_total_samples (int): If sampling, num total samples equal to the number of total anchors; Otherwise, it is the number of positive anchors.

输出

dict[str, Tensor]: A dictionary of loss components.

1、使用交叉熵计算分类损失

        # classification loss
        labels = labels.reshape(-1)
        label_weights = label_weights.reshape(-1)
        cls_score = cls_score.permute(0, 2, 3,1).reshape(-1, self.cls_out_channels)
        loss_cls = self.loss_cls(cls_score, labels, label_weights, avg_factor=num_total_samples)

1、使用Smooth L1计算边界框回归损失

        # regression loss
        bbox_targets = bbox_targets.reshape(-1, 4)
        bbox_weights = bbox_weights.reshape(-1, 4)
        bbox_pred = bbox_pred.permute(0, 2, 3, 1).reshape(-1, 4)
        if self.reg_decoded_bbox:
            # When the regression loss (e.g. `IouLoss`, `GIouLoss`)
            # is applied directly on the decoded bounding boxes, it
            # decodes the already encoded coordinates to absolute format.
            anchors = anchors.reshape(-1, 4)
            bbox_pred = self.bbox_coder.decode(anchors, bbox_pred)
        loss_bbox = self.loss_bbox(
            bbox_pred,
            bbox_targets,
            bbox_weights,
            avg_factor=num_total_samples)