Article
Computer Science, Artificial Intelligence
Niclas Joswig, Juuso Autiosalo, Laura Ruotsalainen
Summary: In this work, a deep learning-based depth estimation model is proposed for real-world bird's-eye perspective in an industry environment with minimal visual cues. The model incorporates the temporal domain for structure from motion estimation and uses a novel architecture built upon the structure from motion learner. The proposed method outperforms the state of the art in difficult context-sparse environments, as shown by evaluation with ground truth depth.
MACHINE VISION AND APPLICATIONS
(2023)
Article
Agriculture, Multidisciplinary
Lehao Tan, Lei Zhou, Nan Zhao, Yong He, Zhengjun Qiu
Summary: A ultra-portable SPAD evaluation system, SPAD-Cap, was proposed for leaf SPAD distribution analysis using an integrated RGB camera module, Raspberry Pi Zero, and light source. Two modeling methods, PLSR and CNN-R, were used to evaluate pixel level SPAD value based on color features. The SPAD distribution map was generated by traversing all pixels of the tested blade sample using the trained model.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2021)
Review
Fisheries
Daoliang Li, Qi Wang, Xin Li, Meilin Niu, He Wang, Chunhong Liu
Summary: This paper provides an overview of machine vision models applied in fish classification, discusses specific applications of various classification methods, and explores the challenges and future research directions in the field of fish classification.
ICES JOURNAL OF MARINE SCIENCE
(2022)
Article
Physics, Multidisciplinary
Ching-Hsun Tseng, Shin-Jye Lee, Jianan Feng, Shengzhong Mao, Yu-Ping Wu, Jia-Yu Shang, Xiao-Jun Zeng
Summary: This work proposes an efficient and robust backbone, UPANets, which utilizes channel and spatial direction attentions to expand the receptive fields in shallow convolutional layers. Experimental results show that UPANets achieve better performance with fewer resources on CIFAR-{10, 100} than existing state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Aisha Batool, Muhammad Wasif Nisar, Muhammad Attique Khan, Jamal Hussain Shah, Usman Tariq, Robertas Damasevicius
Summary: The development of AI applications, particularly intelligent visual aid for Traffic Sign Recognition in Intelligent Transport Systems, has significantly reshaped and transformed the control and safety of automobiles. This study proposes a novel 30-layer deep Convolutional Neural Network model that can robustly and efficiently recognize traffic signs in challenging environmental conditions.
IMAGE AND VISION COMPUTING
(2023)
Article
Ecology
Jarrett Blair, Michael D. Weiser, Kirsten de Beurs, Michael Kaspari, Cameron Siler, Katie E. Marshall
Summary: This study presents a practical methodology of using machine learning in ecological data acquisition pipelines, training and testing algorithms to classify a large number of terrestrial invertebrate specimens. The study addresses issues of inconsistent taxonomic label specificity and unknown taxa classification. The results show that complex machine learning methods are not necessarily more accurate than traditional methods, and the inclusion of contextual metadata improves accuracy.
ECOLOGICAL INFORMATICS
(2022)
Article
Environmental Sciences
Carmine Massarelli, Claudia Campanale, Vito Felice Uricchio
Summary: In this study, a Computer Vision and Machine-Learning-based system was developed to quickly and automatically count and classify microplastics. The machine learning algorithm, supervised classification, and unsupervised classification methods were utilized to determine microplastic quantities, properties, and hidden information. These methods show promise in providing a reliable automated approach for microplastic quantification with significant prospects in method standardisation.
Article
Ecology
Sarkhan Badirli, Christine Johanna Picard, George Mohler, Frannie Richert, Zeynep Akata, Murat Dundar
Summary: Machine learning can be used to create an accurate and efficient method for classifying insect species, including both described and undescribed species. A deep hierarchical Bayesian model is proposed, which can classify samples based on the taxonomic hierarchy of insects. The combination of image and DNA data in the model leads to significant improvement in classification accuracy.
METHODS IN ECOLOGY AND EVOLUTION
(2023)
Article
Engineering, Biomedical
Salih Ertug Ovur, Xuanyi Zhou, Wen Qi, Longbin Zhang, Yingbai Hu, Hang Su, Giancarlo Ferrigno, Elena De Momi
Summary: A novel autonomous learning framework is proposed in this paper to integrate the benefits of both depth vision and EMG signals, achieving real-time hand gesture recognition. Experimental results demonstrate prominent performance by introducing depth information for real-time data labeling.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2021)
Article
Computer Science, Information Systems
Chiranjibi Sitaula, Tej Bahadur Shahi, Faezeh Marzbanrad, Jagannath Aryal
Summary: With the rise of deep learning algorithms, scene image representation methods have improved significantly in accuracy for classification. However, the complexity of scene images leads to intra-class dissimilarity and inter-class similarity problems, limiting the overall performance. This paper reviews existing methods and compares their performance qualitatively and quantitatively, while also speculating on future research directions. This survey provides in-depth insights and applications of recent scene image representation methods.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Partha Das, Sezer Karaoglu, Theo Gevers
Summary: This study explores the intrinsic image decomposition problem, introducing physics-based priors and a new architecture to address it, and evaluates and compares the method on synthetic and real-world datasets.
COMPUTER VISION AND IMAGE UNDERSTANDING
(2022)
Article
Automation & Control Systems
Altaf Hussain, Samee Ullah Khan, Noman Khan, Mohammad Shabaz, Sung Wook Baik
Summary: The integration of artificial intelligence (AI) into human activity recognition (HAR) in smart surveillance systems has the potential to revolutionize behavior monitoring, improving security and surveillance measures. A proposed AI-based behavior biometrics framework is introduced, utilizing a dynamic attention fusion unit (DAFU) and temporal-spatial fusion (TSF) network to effectively recognize human activity in surveillance systems.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2024)
Article
Biology
Amna Khan, Shahzad Rasool
Summary: This paper uses machine learning algorithms to analyze players' EEG data in different types of video games to understand the influence of emotions and personality traits on game choices. The results show that a single-channel EEG device can classify four discrete emotions, and extroversion and neuroticism can affect players' game preferences.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Environmental Sciences
Samantha Phan, Diego Torrejon, Jordan Furseth, Erin Mee, Christine Luscombe
Summary: This paper presents a novel approach that compares unsupervised, weakly-supervised, and supervised methods to facilitate the analysis of microplastics without human-labeled data. The results show that the weakly-supervised approach outperforms the unsupervised and supervised methods in segmentation and classification tasks. The study also demonstrates the potential for automating the identification of microplastics based on their morphology.
SCIENCE OF THE TOTAL ENVIRONMENT
(2023)
Article
Computer Science, Artificial Intelligence
Patrick B. Rodrigues, Yijing Xiao, Yoko E. Fukumura, Mohamad Awada, Ashrant Aryal, Burcin Becerik-Gerber, Gale Lucas, Shawn C. Roll
Summary: Sedentary activity and static postures are associated with work-related musculoskeletal disorders. This study proposes an automated ergonomic assessment algorithm, 3D-AJA, using a Microsoft Kinect camera and open-source computer vision algorithms to monitor office worker postures. The results show that 3D-AJA has a relatively good performance in evaluating joint angles and classifying RULA scores, even with occlusion on lower limbs.
ADVANCED ENGINEERING INFORMATICS
(2022)