Article
Psychology, Multidisciplinary
Ruth Kimchi, Dina Devyatko, Shahar Sabary
Summary: This study examined the role of visual awareness in amodal completion, specifically comparing the effect on local versus global completion. Participants were presented with partially occluded shapes and were asked to discriminate the shape of a target. The results suggest that local completion can occur without visual awareness, but only when the visible occluded shape generates a single, local completion. No completion, either local or global, appears to take place in the absence of visual awareness when the visible occluded shape generates multiple completions. These findings provide insight into the differential role of visual awareness in local and global completions.
FRONTIERS IN PSYCHOLOGY
(2023)
Article
Computer Science, Software Engineering
Bo Gao, Michael W. Spratling
Summary: This paper proposes a novel tracking architecture that increases robustness by considering both the appearance of the tracked object and the appearance of detected distractors in previous frames using explaining away inference. The proposed method, when combined with various existing trackers, improves tracking accuracy and achieves competitive performance on popular benchmarks.
Article
Engineering, Biomedical
Shishun Tian, Minghuo Zheng, Wenbin Zou, Xia Li, Lu Zhang
Summary: This paper proposes a novel system to assist visually impaired individuals in understanding dynamic crosswalk scenes, providing indications of when and where to cross the road and conveying surrounding scene information through audio signals. Experimental results demonstrate that the system is robust and practical for use.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2021)
Article
Psychology, Experimental
Jason K. Chow, Thomas J. Palmeri, Graham Pluck, Isabel Gauthier
Summary: General object recognition ability predicts performance across various visual tests, categories, and haptic recognition. Auditory object recognition ability is highly correlated with visual object recognition ability, indicating a domain-general ability that applies across different situations.
Article
Psychology, Multidisciplinary
Yaqi Wang, Wen Ma, Kai Yang
Summary: This study examined the role of recognizability in modulating the scene consistency effect on object and scene recognition. The results showed that contextual consistency affected the recognition accuracy of both objects and scenes, with a stronger effect on less recognizable stimuli. The mediating role of recognizability was larger for object recognition than for scene recognition. Additionally, contextual inconsistency interfered with both object and scene recognition, as shown in the comparison with other experiments.
CURRENT PSYCHOLOGY
(2023)
Article
Neurosciences
Elissa M. Aminoff, Tess Durham
Summary: Objects are crucial for understanding scenes, and the processing of scenes in the brain is often discussed in contrast to the processing of objects. Using functional magnetic resonance imaging, this study found a significant correlation between the objects within a scene and the neural representation of scenes, particularly in the scene-preferring regions of the brain. These findings indicate that visual processing regions are better characterized by the processes involved when interacting with the stimulus kind rather than the stimulus kind itself.
Article
Computer Science, Information Systems
Yanhua Shao, Xiao Zhang, Kuisheng Liao, Hongyu Chu
Summary: Unmanned aerial vehicle (UAV) based aerial visual tracking is a research hotspot, but mainstream UAV trackers have two shortcomings: accuracy-speed trade-off and restriction from object occlusion and camera motion. To address these, a Fast-AutoTrack tracker based on scene-perceptual memory is proposed, which uses a confidence score to perceive and judge tracking anomalies and predicts search regions for object re-detection. The model updating is accelerated using the perceptual hashing algorithm. On aerial tracking datasets, Fast-AutoTrack outperforms AutoTrack in terms of speed while maintaining similar accuracy, demonstrating a favorable accuracy-speed trade-off.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2023)
Article
Computer Science, Artificial Intelligence
Mridul Ghosh, Himadri Mukherjee, Sk Md Obaidullah, Xiao-Zhi Gao, Kaushik Roy
Summary: Computational perception has experienced a significant transformation from handcrafted feature-based techniques to deep learning in the field of scene text identification and recognition. Over the past decade, there have been important developments and advancements in this area. The traditional handcrafted feature-based techniques have been replaced by deep learning-based techniques, leading to a new stage in scene text identification.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Computer Science, Artificial Intelligence
Nikita Dvornik, Julien Mairal, Cordelia Schmid
Summary: Data augmentation is crucial for training visual recognition systems, helping to reduce overfitting and improve generalization by artificially increasing training examples. In tasks such as object detection, semantic and instance segmentation, augmenting training images by blending objects in existing scenes can significantly enhance model performance.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Ahmad Jalal, Abrar Ahmed, Adnan Ahmed Rafique, Kibum Kim
Summary: The study introduces a novel scene semantic recognition framework, which intelligently segments object locations, generates a novel Bag of Features, and utilizes Maximum Entropy for scene recognition. During experiments, the system demonstrated high accuracy rates on various datasets and shows promising applications in different fields.
Article
Multidisciplinary Sciences
Magdalena Szubielska, Marta Szewczyk, Pawel Augustynowicz, Wojciech Kedziora, Wenke Moehring
Summary: The present study examined adults' strategies of spatial scaling from memory in visual, haptic, and visuo-haptic conditions. The findings suggest that adults consistently employ mental transformation strategies for spatial scaling, regardless of perceptual modality and scaling direction.
SCIENTIFIC REPORTS
(2023)
Article
Computer Science, Information Systems
Lin Guo, Guoliang Fan
Summary: This study proposes a new method flow for instance-level object detection in indoor scenes utilizing pixel-level labeling information, aiming to integrate semantic labeling and instance segmentation for comprehensive understanding. By optimizing instance segmentation through considering spatial fitness and relational context encoded by three graphical models, the method shows significant improvement in small object segmentation according to experimental results.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Xing Zhang, Zuxuan Wu, Yu-Gang Jiang
Summary: Video recognition aims to understand the semantic contents involving interactions between humans and related objects in specific scenes. The fusion of object, scene, and action features is commonly used to improve recognition accuracy. In this paper, the authors propose a method that breaks down the fusion of three features into two pairwise feature relation modeling processes, which helps overcome the challenge of correlation learning in high dimensional features. The proposed method achieves better results with less computational effort compared to alternative methods.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Chemistry, Analytical
Shuhua Liu, Huixin Xu, Qi Li, Fei Zhang, Kun Hou
Summary: This paper presents an object recognition method based on scene text reading, which improves text detection accuracy and recognition accuracy through deep learning models and dataset training, effectively addressing the issue of robot object recognition in complex scenes.
Article
Computer Science, Information Systems
Kitsaphon Thitisiriwech, Teerapong Panboonyuen, Pittipol Kantavat, Yuji Iwahori, Boonserm Kijsirikul
Summary: This paper presents a semantic segmentation method for autonomous driving systems, which enhances the DeepLab-V3-A1 architecture by modifying the Xception model and utilizing a different number of 1x1 convolution layers. Experimental results show that our proposed method performs comparably to the baseline methods on various measurement units. Additionally, we contribute the Bangkok Urbanscapes dataset, aiming to improve autonomous driving systems in cities with unique traffic and driving conditions.