基于分类和回归子网的包围盒预测（siamGAT和siamCAR的头部）

最新推荐文章于 2024-07-03 14:43:18 发布

奋斗的ran雪

最新推荐文章于 2024-07-03 14:43:18 发布

阅读量878

点赞数 3

分类专栏：论文学习文章标签：深度学习神经网络

本文链接：https://blog.csdn.net/every_step/article/details/120760749

版权

论文学习专栏收录该内容

20 篇文章 4 订阅

订阅专栏

本文档详细介绍了SiamGAT和SiamCAR预测模块的损失函数实现，该函数通过计算交并比（IoU）来衡量预测框与目标框的匹配程度。代码中展示了如何计算两框的交集面积、并集面积以及损失，用于训练模型进行精确的目标定位。

摘要由CSDN通过智能技术生成

siamGAT和siamCAR都使用相同的预测模块，具体参考siamCAR论文原文：https://arxiv.org/abs/1911.07241

3.2. Bounding Box Prediction with Classification
and Regression Subnetwork

在代码中实现为

    def forward(self, pred, target, weight=None):
        pred_left = pred[:, 0]
        pred_top = pred[:, 1]
        pred_right = pred[:, 2]
        pred_bottom = pred[:, 3]

        target_left = target[:, 0]
        target_top = target[:, 1]
        target_right = target[:, 2]
        target_bottom = target[:, 3]

        target_aera = (target_left + target_right) * \
                      (target_top + target_bottom)
        pred_aera = (pred_left + pred_right) * \
                    (pred_top + pred_bottom)

        w_intersect = torch.min(pred_left, target_left) + \
                      torch.min(pred_right, target_right)
        h_intersect = torch.min(pred_bottom, target_bottom) + \
                      torch.min(pred_top, target_top)

        area_intersect = w_intersect * h_intersect
        area_union = target_aera + pred_aera - area_intersect

        losses = -torch.log((area_intersect + 1.0) / (area_union + 1.0))

        if weight is not None and weight.sum() > 0:
            return (losses * weight).sum() / weight.sum()
        else:
            assert losses.numel() != 0
            return losses.mean()