Camouflaged objects are generally difficult to be detected in their natural environment even for human beings. In this paper, we propose a novel bio-inspired network, named the MirrorNet, that leverages both instance segmentation and mirror stream camouflaged object segmentation. Differently from existing networks segmentation, our proposed network possesses two streams: main corresponding with original image its flipped image, respectively. The output is then fused into stream's result...

Abstract Modern malware evolves various detection avoidance techniques to bypass the state‐of‐the‐art methods. An emerging trend deal with this issue is combination of image transformation and machine learning models classify detect malware. However, existing works in field only perform simple These transformations have not considered color encoding pixel rendering on performance classifiers. In article, we propose a novel approach arranging bytes from binary files into images. developed...

Human action recognition is valuable for numerous practical applications, e.g., gaming, video surveillance, and search. In this paper we hypothesize that the classification of actions can be boosted by designing a smart feature pooling strategy under prevalently used bag-of-words-based representation. Founded on automatic saliency analysis, propose spatial-temporal attention-aware scheme pooling. First, saliencies are predicted using model, localized features pooled at different levels...

Modern malware evolves various detection avoidance techniques to bypass the state-of-the-art methods. An emerging trend deal with this issue is combination of image transformation and machine learning classify detect malware. However, existing works in field only perform simple methods that limit accuracy detection. In paper, we introduce a novel approach by using deep network on images transformed from binary samples. particular, first develop hybrid method convert binaries into color...

In this demo, we present a practical system, magic closet, for automatic occasion-oriented clothing pairing. Given user-input occasion, e.g., wedding or shopping, the closet intelligently and automatically pairs user-specified reference (upper-body lower-body) with most suitable one from online shops. Two key criteria are explicitly considered system. One criterion is to wear properly, compared suit pants, it more decent cocktail dress banquet occasion. The other aesthetically, red T-shirt...

Recently visual saliency has attracted wide attention of researchers in the computer vision and multimedia field. However, most saliency-related research was conducted on still images for studying static saliency. In this paper, we give a comprehensive comparative study first time dynamic (video shots) (key frames corresponding video shots), two key observations are obtained: 1) is often different from, yet quite related with, image saliency, 2) camera motions, such as tilting, panning or...

This paper pushes the envelope on decomposing camouflaged regions in an image into meaningful components, namely, instances. To promote new task of instance segmentation in-the-wild images, we introduce a dataset, dubbed CAMO++, that extends our preliminary CAMO dataset (camouflaged object segmentation) terms quantity and diversity. The substantially increases number images with hierarchical pixel-wise ground truths. We also provide benchmark suite for segmentation. In particular, present...

In this article, we adopt the maximizing mutual information (MI) approach to tackle problem of unsupervised learning binary hash codes for efficient cross-modal retrieval. We proposed a novel method, dubbed info-max hashing (CMIMH). First, learn informative representations that can preserve both intramodal and intermodal similarities, leverage recent advances in estimating variational lower bound MI between input features different modalities. By jointly these MIs under assumption are...

Anomaly detection is an area of video analysis and plays increasing role in ensuring safety, preventing risks, guaranteeing quick response intelligent surveillance systems. It has become a popular research topic piqued the interest researchers different communities, such as computer vision, machine learning, remote sensing, data mining, recent years. This promotes novel mobile systems where drones are equipped with cameras to help people find better more efficient solutions automatically...

Discovering the secret of beauty has been pursuit artists and philosophers for centuries. Nowadays, computational model estimation actively explored in computer science community, yet with focus mainly on facial features. In this work, we perform a comprehensive study female attractiveness conveyed by single/multiple modalities cues, i.e., face, dressing and/or voice, aim to uncover how different individually collectively affect human sense beauty. To end, collect first Multi-Modality Beauty...

Anomaly detection plays an increasingly important role in video surveillance and is one of the issues that has attracted various communities, such as computer vision, machine learning, data mining recent years. Moreover, drones equipped with cameras have quickly been deployed to a wide range applications, starting from border security applications street monitoring systems. However, there notable lack adequate drone-based datasets available detect unusual events urban traffic environment,...

The proliferation of Artificial Intelligence (AI) models such as Generative Adversarial Networks (GANs) has shown impressive success in image synthesis. GAN-based synthesized images have been widely spread over the Internet with advancement generating naturalistic and photo-realistic images. This might ability to improve content media; however, it also constitutes a threat regard legitimacy, authenticity, security. Moreover, implementing an automated system that is able detect recognize...

Camouflaged object detection and segmentation is a new challenging research topic in computer vision. There serious issue of lacking data on concealed objects such as camouflaged animals natural scenes. In this paper, we address the problem few-shot learning for segmentation. To end, first collect dataset, CAMO-FS, benchmark. As instances are to recognize due their similarity compared surroundings, guide our models obtain features that highly distinguish from background. work, propose...

In this paper, we propose a computational framework, called Image Re-Attentionizing, to endow the target region in an image with ability of attracting human visual attention. particular, objective is recolor patches by color transfer naturalness and smoothness preserved yet attention augmented. We approach within Markov Random Field (MRF) framework extended graph cuts method developed pursue solution. The input first over-segmented into patches, as well their neighbors are used construct...

Document Imaging Understanding (DIU) is the process of converting all information content a document image digital into an electronic format launched its reasonable content. We first evaluate different state-of-the-art object detection methods (Fast R-CNN and Faster R-CNN) for task on POD dataset. observe that each method sensitive to certain objects. Therefore, we propose combining detections Fast RCNN in order exploit advantages two models. Through extensive experiments, our proposed...

Unmanned aircraft systems or drones enable us to record capture many scenes from the bird’s-eye view and they have been fast deployed a wide range of practical domains, i.e., agriculture, aerial photography, delivery surveillance. Object detection task is one core steps in understanding videos collected drones. However, this very challenging due unconstrained viewpoints low resolution captured videos. While deep-learning modern object detectors recently achieved great success general...