Video Object Segmentation Using Space-Time Memory Networks

Abstract

We propose a novel solution for semi-supervised video object segmentation. By the nature of the problem, available cues (e.g. video frame(s) with object masks) become richer with the intermediate predictions. However, the existing methods are unable to fully exploit this rich source of information. We resolve the issue by leveraging memory networks and learn to read relevant information from all available sources. In our framework, the past frames with object masks form an external memory, and the current frame as the query is segmented using the mask information in the memory. Specifically, the query and the memory are densely matched in the feature space, covering all the space-time pixel locations in a feed-forward fashion. Contrast to the previous approaches, the abundant use of the guidance information allows us to better handle the challenges such as appearance changes and occlussions. We validate our method on the latest benchmark sets and achieved the state-of-the-art performance (overall score of 79.4 on Youtube-VOS val set, J of 88.7 and 79.2 on DAVIS 2016/2017 val set respectively) while having a fast runtime (0.16 second/frame on DAVIS 2016 val set).

Learning What to Learn for Video Object Segmentation

“…However, this strategy easily leads to overfitting to the initial target appearance and impractically long run-times. More recent methods [34,13,32,23,36,24,17] therefore integrate target-specific appearance models into the segmentation architecture. In addition to improved run-times, many of these methods can also benefit from full end-to-end learning, which has been shown to have a crucial impact on performance [32,14,24].…”

“…While most state-of-the-art VOS approaches employ similar image feature extractors and segmentation heads, the advances in how to capture and utilize target information has led to much improved performance [14,32,24,28]. A Fig.…”

“…promising direction is to employ feature matching techniques [13,14,32,24] in order to compare the reference frame with new images to segment. Such feature feature-matching layers greatly benefit from their efficiency and differentiability.…”

“…Such feature feature-matching layers greatly benefit from their efficiency and differentiability. This allows the design of fully end-to-end trainable architectures, which has been shown to be important for segmentation performance [14,32,24]. On the other hand, feature matching relies on a powerful and generic feature embedding, which may limit its performance in challenging scenarios.…”

Overview of deep-learning based methods for salient object detection in videos

“…Inspired by recent advances in related video tasks, e.g., dense connections for spatio-temporal interaction in action recognition [39] and space-time memory block in video object segmentation [40], we possibly consider these techniques to avoid the latent dependency issues in video SOD. Learning spatial-temporal features in an end-to-end manner is important for further accuracy improvement.…”