Search

Your search keyword '"Gulyanon, Sarun"' showing total 13 results
Start Over Author "Gulyanon, Sarun"
13 results on '"Gulyanon, Sarun"'

2. VISHIEN-MAAT: Scrollytelling visualization design for explaining Siamese Neural Network concept to non-technical users

Author

Chotisarn, Noptanit, Gulyanon, Sarun, Zhang, Tianye, and Chen, Wei

Subjects
Computer Science - Human-Computer Interaction, Computer Science - Machine Learning
Abstract

The past decade has witnessed rapid progress in AI research since the breakthrough in deep learning. AI technology has been applied in almost every field; therefore, technical and non-technical end-users must understand these technologies to exploit them. However existing materials are designed for experts, but non-technical users need appealing materials that deliver complex ideas in easy-to-follow steps. One notable tool that fits such a profile is scrollytelling, an approach to storytelling that provides readers with a natural and rich experience at the reader's pace, along with in-depth interactive explanations of complex concepts. Hence, this work proposes a novel visualization design for creating a scrollytelling that can effectively explain an AI concept to non-technical users. As a demonstration of our design, we created a scrollytelling to explain the Siamese Neural Network for the visual similarity matching problem. Our approach helps create a visualization valuable for a short-timeline situation like a sales pitch. The results show that the visualization based on our novel design helps improve non-technical users' perception and machine learning concept knowledge acquisition compared to traditional materials like online articles.

Published
2023
Full Text
View/download PDF

3. Data Generation for Satellite Image Classification Using Self-Supervised Representation Learning

Author

Gulyanon, Sarun, Limprasert, Wasit, Songmuang, Pokpong, and Kongkachandra, Rachada

Subjects
Computer Science - Computer Vision and Pattern Recognition, 68T45, I.2.10
Abstract

Supervised deep neural networks are the-state-of-the-art for many tasks in the remote sensing domain, against the fact that such techniques require the dataset consisting of pairs of input and label, which are rare and expensive to collect in term of both manpower and resources. On the other hand, there are abundance of raw satellite images available both for commercial and academic purposes. Hence, in this work, we tackle the insufficient labeled data problem in satellite image classification task by introducing the process based on the self-supervised learning technique to create the synthetic labels for satellite image patches. These synthetic labels can be used as the training dataset for the existing supervised learning techniques. In our experiments, we show that the models trained on the synthetic labels give similar performance to the models trained on the real labels. And in the process of creating the synthetic labels, we also obtain the visual representation vectors that are versatile and knowledge transferable., Comment: 11 pages, 6 figures, 5 tables. Submitted to Science & Technology Asia

Published
2022

8. A Comparative Study of Noise Augmentation and Deep Learning Methods on Raman Spectral Classification of Contamination in Hard Disk Drive

Author

Gulyanon, Sarun, primary, Deepaisam, Somrudee, additional, Srisumarnk, Chayud, additional, Chiewnawintawat, Nattapol, additional, Anzkoonsawaenasuk, Angkoon, additional, Laitrakun, Seksan, additional, Ooaorakasit, Pakorn, additional, Rakpongsiri, Pornchai, additional, Meechamnan, Thawanpat, additional, and Sompongse, Duangporn, additional

Published
2022
Full Text
View/download PDF

10. Direction Selectivity in Drosophila Proprioceptors Requires the Mechanosensory Channel Tmc

Author

He, Liping, Gulyanon, Sarun, Skanata, Mirna Mihovilovic, Karagyozov, Doycho, Heckscher, Ellie S., Krieg, Michae, Tsechpenakis, Gavriil, Gershow, Marc, W. Daniel, Tracey, Jr, He, Liping, Gulyanon, Sarun, Skanata, Mirna Mihovilovic, Karagyozov, Doycho, Heckscher, Ellie S., Krieg, Michae, Tsechpenakis, Gavriil, Gershow, Marc, and W. Daniel, Tracey, Jr

Abstract

Drosophila Transmembrane channel-like (Tmc) is a protein that functions in larval proprioception. The closely related TMC1 protein is required for mammalian hearing and is a pore-forming subunit of the hair cell mechanotransduction channel. In hair cells, TMC1 is gated by small deflections of microvilli that produce tension on extracellular tip-links that connect adjacent villi. How Tmc might be gated in larval proprioceptors, which are neurons having a morphology that is completely distinct from hair cells, is unknown. Here, we have used high-speed confocal microscopy both to measure displacements of proprioceptive sensory dendrites during larval movement and to optically measure neural activity of the moving proprioceptors. Unexpectedly, the pattern of dendrite deformation for distinct neurons was unique and differed depending on the direction of locomotion: ddaE neuron dendrites were strongly curved by forward locomotion, while the dendrites of ddaD were more strongly deformed by backward locomotion. Furthermore, GCaMP6f calcium signals recorded in the proprioceptive neurons during locomotion indicated tuning to the direction of movement. ddaE showed strong activation during forward locomotion, while ddaD showed responses that were strongest during backward locomotion. Peripheral proprioceptive neurons in animals mutant for Tmc showed a near-complete loss of movement related calcium signals. As the strength of the responses of wild-type animals was correlated with dendrite curvature, we propose that Tmc channels may be activated by membrane curvature in dendrites that are exposed to strain. Our findings begin to explain how distinct cellular systems rely on a common molecular pathway for mechanosensory responses., Peer Reviewed, Postprint (published version)

Published
2019

11. Model and Appearance Based Analysis of Neuronal Morphology from Different Microscopy Imaging Modalities

Author

Gulyanon, Sarun

Subjects
nervous system
Abstract

The neuronal morphology analysis is key for understanding how a brain works. This process requires the neuron imaging system with single-cell resolution; however, there is no feasible system for the human brain. Fortunately, the knowledge can be inferred from the model organism, Drosophila melanogaster, to the human system. This dissertation explores the morphology analysis of Drosophila larvae at single-cell resolution in static images and image sequences, as well as multiple microscopy imaging modalities. Our contributions are on both computational methods for morphology quantification and analysis of the influence of the anatomical aspect. We develop novel model-and-appearance-based methods for morphology quantification and illustrate their significance in three neuroscience studies. Modeling of the structure and dynamics of neuronal circuits creates understanding about how connectivity patterns are formed within a motor circuit and determining whether the connectivity map of neurons can be deduced by estimations of neuronal morphology. To address this problem, we study both boundary-based and centerline-based approaches for neuron reconstruction in static volumes. Neuronal mechanisms are related to the morphology dynamics; so the patterns of neuronal morphology changes are analyzed along with other aspects. In this case, the relationship between neuronal activity and morphology dynamics is explored to analyze locomotion procedures. Our tracking method models the morphology dynamics in the calcium image sequence designed for detecting neuronal activity. It follows the local-to-global design to handle calcium imaging issues and neuronal movement characteristics. Lastly, modeling the link between structural and functional development depicts the correlation between neuron growth and protein interactions. This requires the morphology analysis of different imaging modalities. It can be solved using the part-wise volume segmentation with artificial templates, the standardized representation of neurons. Our method follows the global-to-local approach to solve both part-wise segmentation and registration across modalities. Our methods address common issues in automated morphology analysis from extracting morphological features to tracking neurons, as well as mapping neurons across imaging modalities. The quantitative analysis delivered by our techniques enables a number of new applications and visualizations for advancing the investigation of phenomena in the nervous system.

Published
2018

Catalog

Books, media, physical & digital resources
See catalog results