Unsupervised Metric Relocalization Using Transform Consistency Loss
Unsupervised Metric Relocalization Using Transform Consistency Loss
复制标题
DOI：
--
复制
发表时间：
2020-11
期刊：
影响因子：
--
通讯作者：
Mike Kasper;Fernando Nobre;C. Heckman;Nima Keivan
中科院分区：
文献类型：
--
作者：
Mike Kasper;Fernando Nobre;C. Heckman;Nima Keivan
传统上，训练网络进行度量重定位需要准确的图像对应。在实践中，这些是通过限制域覆盖、使用额外传感器或捕获大型多视角数据集来获得的。相反，我们提出了一个自我监督的解决方案，它利用了一个关键的洞察力：在地图中定位查询图像应该产生相同的绝对姿势，而不考虑用于配准的参考图像。在这种直觉的指导下，我们推导出了一种新的变换一致性损失。利用这个损失函数，我们训练一个深度神经网络来推断密集特征图和显著图，从而在动态环境中执行稳健的度量重定位。我们在人工合成数据和真实世界数据上对我们的框架进行了评估，结果表明，在有限数量的地面真实信息可用时，我们的方法优于其他监督方法。
Training networks to perform metric relocalization traditionally requires accurate image correspondences. In practice, these are obtained by restricting domain coverage, employing additional sensors, or capturing large multi-view datasets. We instead propose a self-supervised solution, which exploits a key insight: localizing a query image within a map should yield the same absolute pose, regardless of the reference image used for registration. Guided by this intuition, we derive a novel transform consistency loss. Using this loss function, we train a deep neural network to infer dense feature and saliency maps to perform robust metric relocalization in dynamic environments. We evaluate our framework on synthetic and real-world data, showing our approach outperforms other supervised methods when a limited amount of ground-truth information is available.