Self-supervised Sparse to Dense Motion Segmentation
Self-supervised Sparse to Dense Motion Segmentation
复制标题
DOI：
10.1007/978-3-030-69532-3_26
复制
发表时间：
2020-08
期刊：
影响因子：
--
通讯作者：
Amirhossein Kardoost;Kalun Ho;Peter Ochs;M. Keuper
中科院分区：
文献类型：
--
作者：
Amirhossein Kardoost;Kalun Ho;Peter Ochs;M. Keuper
视频中的可观察运动可以产生相对于场景移动的对象的定义。分割这样的移动对象的任务被称为运动分割，通常通过在长的稀疏点轨迹中聚合运动信息来解决，或者通过依赖于大量训练数据直接产生每帧密集分割来解决。在本文中，我们提出了一种自监督的方法来学习从单个视频帧的稀疏运动分割的致密化。虽然以前的运动分割方法建立在大型代理数据集的预训练基础上，并使用密集的运动信息作为像素分割的基本线索，但我们的模型不需要预训练，并在测试时对单帧进行操作。它可以以特定于序列的方式进行训练，以从稀疏和嘈杂的输入中产生高质量的密集分割。我们在著名的运动分割数据集FBMS 59和DAVIS 2016上评估了我们的方法。
Observable motion in videos can give rise to the definition of objects moving with respect to the scene. The task of segmenting such moving objects is referred to as motion segmentationand is usually tackled either by aggregating motion information in long, sparse point trajectories, or by directly producing per frame dense segmentations relying on large amounts of training data. In this paper, we propose a self supervised method to learn the densification of sparse motion segmentations from single video frames. While previous approaches towards motion segmentation build upon pre-training on large surrogate datasets and use dense motion information as an essential cue for the pixelwise segmentation, our model does not require pre-training and operates at test time on single frames. It can be trained in a sequence specific way to produce high quality dense segmentations from sparse and noisy input. We evaluate our method on the well-known motion segmentation datasets FBMS59 and DAVIS2016.