Rapid dynamical pattern recognition for sampling sequences

Wu, Weiming; Wang, Qian; Yuan, Chengzhi

doi:10.1007/s11432-019-2878-y

Cited by 13 publications

(3 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Training. A meta-training method is used for learning the meta knowledge ω defined in (1), where the training data consist of a number of different source tasks {D Support Source ∪ D Query Source } i ∈ T Source . Following the "episodic training" procedure in [47], we use only one task for each iteration.…”

Section: Implementation Detailsmentioning

confidence: 99%

“…Video object segmentation (VOS) aims to separate foreground objects from the background in a video sequence and is a fundamental task in computer vision; it has important applications, such as video detection [1], classification [2], and reconstruction [3]. Typically, according to whether or not annotations are provided for the first frame during testing, VOS can be categorized into few-shot VOS [4,5] and zero-shot VOS [6].…”

Section: Introductionmentioning

confidence: 99%

“…Advances in deep learning and the introduction of the DAVIS 1) and YouTubeVOS 2) challenge have led to significant progress in semi-supervised VOS. Aiming to be target-aware with a limited annotation, quite a few methods rely heavily on first-frame finetuning [4,5,[7][8][9].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fast target-aware learning for few-shot video object segmentation

Chen

Hao

Yang

et al. 2022

Sci. China Inf. Sci.

View full text Add to dashboard Cite

Few-shot video object segmentation (FSVOS) aims to segment a specific object throughout a video sequence when only the first-frame annotation is given. In this study, we develop a fast target-aware learning approach for FSVOS, where the proposed approach adapts to new video sequences from its firstframe annotation through a lightweight procedure. The proposed network comprises two models. First, the meta knowledge model learns the general semantic features for the input video image and up-samples the coarse predicted mask to the original image size. Second, the target model adapts quickly from the limited support set. Concretely, during the online inference for testing the video, we first employ fast optimization techniques to train a powerful target model by minimizing the segmentation error in the first frame and then use it to predict the subsequent frames. During the offline training, we use a bilevel-optimization strategy to mimic the full testing procedure to train the meta knowledge model across multiple video sequences. The proposed method is trained only on an individual public video object segmentation (VOS) benchmark without additional training sets and compared favorably with state-of-the-art methods on DAVIS-2017, with a J &F overall score of 71.6%, and on YouTubeVOS-2018, with a J &F overall score of 75.4%. Meanwhile, a high inference speed of approximately 0.13 s per frame is maintained.

show abstract

Section: Implementation Detailsmentioning

confidence: 99%