Disaster Image Filtering and Summarization Based on Multi-layered Affinity Propagation

Abstract

Abstract-In this paper, a disaster image filtering and summarization (DIFS) framework is proposed based on multilayered affinity propagation. The proposed framework is able to automatically identify and summarize latent semantic themes (scenes) in a disaster topic and filter junk images at the same time. Specifically, the images belonging to a disaster topic are first clustered into different groups based on visual descriptors using affinity propagation (AP). Then the typical instances within each cluster are collected to perform the second-layer clustering for identifying final positive clusters by utilizing both visual and textual similarities concurrently. At both layers, the proposed curve fitting function is applied to select appropriate preference values for the AP algorithm. The experimental results on the real world Flickr data set demonstrate the effectiveness of the proposed framework.

Highly accurate video object identification utilizing hint information

“…Video sequence contains a large number of frames which include intrinsic spatio-temporal information that could be used to extract hint information to help object identification. Although there is some existing work using clustering-based method for selecting representative exemplars from a large corpus of static images [6], effectively extracting useful and compact information from video as a hint to help with object identification is a challenging research problem which has not been deeply explored.…”

Exploring Hidden Coherent Feature Groups and Temporal Semantics for Multimedia Big Data Analysis

“…Because of its simplicity, general applicability, and performance, the affinity propagation (AP) algorithm has found application in the fields of science and engineering [138], which inspires us to adapt it to our framework for feature clustering. Specifically, we choose to use the AP algorithm for the following reasons:…”

“…The work in [138] shows that the number of groups is monotonically increasing with P polynomially. The value of P is empirical set to -10 in the following experiments.…”

“…Then an unsupervised filtering and summarization approach is proposed to automatically identify and summarize latent semantics in a topic and filter irrelevant items at the same time [187]. Finally, a multimedia semantic retrieval system is developed based the proposed framework on mobile devices for further evaluation [181].…”

Integrating Deep Learning with Correlation-based Multimedia Semantic Concept Detection

“…Because multimedia data carries knowledgeable information, it has been widely adopted to different genera [210,210,211,214,217,233,241]. For example, multimedia data is leveraged to combine education with entertainment and pass the knowledge to the next generation in a very joyful way; medical science is always eager to identify the abnormal cases from prolific multimedia content [136]; disaster management organization can provide the service on time if they can effectively analyze the incoming multimedia information [22,63,126,167,182,209,214,216]; and the engineering research area also started the trend of mining multimedia content in building a sustainable building [72,88].…”