The recently proposed camouflaged object detection (COD) attempts to segment objects that are visually blended into their surroundings, which is extremely complex and difficult in real-world scenarios. Apart from high intrinsic similarity between the background, usually diverse scale, fuzzy appearance, even severely occluded. To deal with these problems, we propose a mixed-scale triplet network, Zoom- Net, mimics behavior of humans when observing vague images, i.e., zooming out....

Camouflaged object detection (COD), segmenting objects that are elegantly blended into their surroundings, is a valuable yet challenging task. Existing deep-learning methods often fall the difficulty of accurately identifying camouflaged with complete and fine structure. To this end, in paper, we propose novel boundary-guided network (BGNet) for detection. Our method explores extra object-related edge semantics to guide representation learning COD, which forces model generate features...

Vision transformers have recently shown strong global context modeling capabilities in camouflaged object detection. However, they suffer from two major limitations: less effective locality and insufficient feature aggregation decoders, which are not conducive to camou-flaged detection that explores subtle cues indistinguishable backgrounds. To address these issues, this paper, we propose a novel transformer-based Feature Shrinkage Pyramid Network (FSPNet), aims hierarchically decode...

Recent camouflaged object detection (COD) attempts to segment objects visually blended into their surroundings, which is extremely complex and difficult in real-world scenarios.Apart from the high intrinsic similarity between background, are usually diverse scale, fuzzy appearance, even severely occluded.To this end, we propose an effective unified collaborative pyramid network that mimics human behavior when observing vague images videos, i.e., zooming out.Specifically, our approach employs...

The purpose of co-salient object detection (CoSOD) is to detect the salient objects that co-occur in a group relevant images. CoSOD has been significantly prospered by recent advances convolutional neural networks (CNNs). However, it shows general limitations modeling long-range feature dependencies, which crucial for CoSOD. In vision transformer, self-attention mechanism utilized capture global dependencies but unfortunately destroy local spatial details, are also essential To address above...

Camouflaged object detection (COD) aims to identify the objects that seamlessly blend into surrounding backgrounds. Due intrinsic similarity between camouflaged and background region, it is extremely challenging precisely distinguish by existing approaches. In this paper, we propose a hierarchical graph interaction network termed HGINet for detection, which capable of discovering imperceptible via effective among tokenized features. Specifically, first design region-aware token focusing...

Camouflaged object detection (COD), segmenting objects that are elegantly blended into their surroundings, is a valuable yet challenging task. Existing deep-learning methods often fall the difficulty of accurately identifying camouflaged with complete and fine structure. To this end, in paper, we propose novel boundary-guided network (BGNet) for detection. Our method explores extra object-related edge semantics to guide representation learning COD, which forces model generate features...

Co-salient object detection (CoSOD) aims at detecting common salient objects within a group of relevant source images. Most the latest works employ attention mechanism for finding objects. To achieve accurate CoSOD results with high-quality maps and high efficiency, we propose novel Memory-aided Contrastive Consensus Learning (MCCL) framework, which is capable effectively co-salient in real time (∼150 fps). learn better consensus, Group Aggregation Module (GCAM) to abstract features each...

In this article, we provide a comprehensive study of new task called collaborative camouflaged object detection (CoCOD), which aims to simultaneously detect objects with the same properties from group relevant images. To end, meticulously construct first large-scale dataset, termed CoCOD8K, consists 8528 high-quality and elaborately selected images mask annotations, covering five superclasses 70 subclasses. The dataset spans wide range natural artificial camouflage scenes diverse appearances...

Due to the high similarity between camouflaged instances and background, recently proposed instance segmentation (CIS) faces challenges in accurate localization segmentation. To this end, inspired by query-based transformers, we propose a unified multi-task learning framework for segmentation, termed UQFormer, which builds set of mask queries boundary learn shared composed query representation efficiently integrates global object region cues, simultaneous detection scenarios. Specifically,...

Warping-based image stitching methods often suffer from perspective variations among multiple images and lead to shape distortions in results. Moreover, they also quickly lose their efficiency low-textured images, due the lack of reliable point correspondences. To solve these problems, this paper presents a locally warping-based by imposing line constraints. First, two-stage alignment scheme with constraints is introduced achieve accurate alignment. More precisely, features are adopted as...

Image stitching algorithms often adopt the global transform, such as homography, and work well for planar scenes or parallax free camera motions. However, these conditions are easily violated in practice. With casual motions, variable taken views, large depth change, complex structures, it is a challenging task images. The transform model provides dreadful results, misalignments projective distortions, especially perspective distortion. To this end, we suggest perspective-preserving warping...

Generative (generalized) zero-shot learning [(G)ZSL] models aim to synthesize unseen class features by using only seen feature and attribute pairs as training data. However, the generated fake tend be dominated thus classified classes, which can lead inferior performances under (ZSL), unbalanced results generalized ZSL (GZSL). To address this challenge, we tailor a novel balanced semantic embedding generative network (BSeGN), incorporates into scenarios in pursuit of unbiased GZSL....

Abstract. Image stitching algorithms often adopt the global transform, such as homography, and work well for planar scenes or parallax free camera motions. However, these conditions are easily violated in practice. With casual motions, variable taken views, large depth change, complex structures, it is a challenging task images. The transform model provides dreadful results, misalignments projective distortions, especially perspective distortion. To this end, we suggest...

With the widespread application of artificial intelligence technology in field biomedical images, deep learning-based detection vaginal discharge, an important but challenging topic medical image processing, has drawn increasing amount research interest. Although past few decades have witnessed major advances object natural scenes, such successes been slow to not only because complex background and diverse cell morphology microscope also due scarcity well-annotated datasets objects images....

Camouflaged object detection (COD) aims to segment objects visually embedded in their surroundings, which is a very challenging task due the high similarity between and background. To address it, most methods often incorporate additional information (e.g., boundary, texture, frequency clues) guide feature learning for better detecting camouflaged from Although progress has been made, these are basically individually tailored specific auxiliary cues, thus lacking adaptability not consistently...

Camouflaged object detection (COD) aims to identify the objects that seamlessly blend into surrounding backgrounds. Due intrinsic similarity between camouflaged and background region, it is extremely challenging precisely distinguish by existing approaches. In this paper, we propose a hierarchical graph interaction network termed HGINet for detection, which capable of discovering imperceptible via effective among tokenized features. Specifically, first design region-aware token focusing...