Baolong Liu

This article targets the task of language-based video moment localization. The language-based setting of this task allows for an open set of target activities, resulting in a large variation of the temporal lengths of video moments. Most existing methods prefer to first sample sufficient candidate moments with various temporal lengths, then match them with the given query to determine the target moment. However, candidate moments generated with a fixed temporal granularity may be suboptimal to handle the large variation in moment lengths. To this end, we propose a novel multi-stage Progressive Localization Network (PLN) that progressively localizes the target moment in a coarse-to-fine manner. Specifically, each stage of PLN has a localization branch and focuses on candidate moments that are generated with a specific temporal granularity. The temporal granularities of candidate moments are different across the stages. Moreover, we devise a conditional feature manipulation module and an upsampling connection to bridge the multiple localization branches. In this fashion, the later stages are able to absorb the previously learned information, thus facilitating the more fine-grained localization. Extensive experiments on three public datasets demonstrate the effectiveness of our proposed PLN for language-based moment localization, especially for localizing short moments in long videos.

Ship License Numbers (SLNs) play a significant role in intelligent waterway transportation systems. However, there is a dearth of published papers concerning the location of SLNs. In this paper, we present an effective coarse-to-fine approach for locating various SLNs. The proposed method contains four main steps. First, for ships in the wild, the coarse problem of locating SLNs is posed as the detection of specific text groups. In this phase, the proximity and similarity laws of Gestalt Theory are collaboratively used to create text-group hypotheses and detect the meaningful region groups that may contain text in input images. In the second step, four SLN prior features are first summarized. An SLN prior-based fine location process is then conducted to determine the locations of text candidates that actually contain characters and text, i.e., the candidate SLNs. Two connected component classification and merging algorithms are proposed based on height continuity and vertical/horizontal projection analysis. Third, a connected-component scatter-based fake-SLN elimination algorithm is developed. The accurate positions of the SLNs in the input image are obtained in this stage. Finally, partially located SLNs are compensated by a method that characterizes the color and brightness similarity features of characters in the same SLN. The proposed approach is then applied to two collected datasets. The approach is proved to be effective with an overall ship-level location accuracy of 70.38% and a SLN-level F-measure of 0.642 on 950 images with 1374 labeled SLNs.