Ngoài hoạt động đào tạo, Khoa Công nghệ thông tin và truyền thông còn tham gia nghiên cứu nhiều đề tài khoa học. Phòng thí nghiệm Công nghệ thông tin và truyền thông (ICTLab) là phòng thí nghiệm nghiên cứu CNTT-TT quốc tế hợp tác giữa USTH và các đối tác Việt Nam / Pháp, bao gồm các nhà nghiên cứu đến từ USTH, IOIT (Viện Công nghệ Thông tin), IRD (Institut de Recherche pour le Développement) và Đại học La Rochelle, Pháp.
ICTLab hiện đang thực hiện các chủ đề nghiên cứu sau:
- Machine Learning, Deep Learning and Data Mining
- Image and Speech Processing
- Modeling and Simulation
- Sensor Networks and Embedded Systems
- High Performance Computing
- Health Informatics and BioInformatics
Những dự án hiện tại của ICTLab:
– iDream (Integration of remote sensing and deep learning technologies for evaluating rice health towards smart agriculture, 2022-2024): The project does the Research on nutritional factors affecting rice plant health, thereby building models to predict rice plant health status based on advanced remote sensing and deep learning technologies. The specific targets are (1) Building a database on rice plant health, including nutrients (N, P, K), minerals (Si), chlorophyll in leaves, multispectral and hyper-spectral images taken from drones and reflectance spectra measured from handheld devices; (2) Building a model to determine the excess and deficiency of nutrients N, P, K in rice fields using multispectral and hyperspectral remote sensing images; (3) Determination of correlations between rice genes and Si mineral uptake capacity using sequenced rice genome datasets and directly sampled rice nutrient datasets and inferred from reflectance spectral data; and (4) Building software to support rice plant health assessment using established models and prediction methods.
– MiJVeh (Mitigation against Jamming Attacks in Vehicular Networks, from 2021-2023): Vehicular Ad hoc Networks (VANETs) plays a role of wireless communication technology supporting Intelligent Transportation System (ITS) in the domain of vehicles. Characteristics of the vehicular environment always create challenges. A wide range of applications and usage scenarios requires enhancements in standardized protocols and security in vehicular networks before implementing in daily life. One of the common types of attacks in VANETs is jamming. Although several detection methods have been proposed to deal with jamming attacks, such methods specified for basic safety messages in vehicular networks still remains an open topic for researching due to the variety of ITS applications and scenarios. Moreover, after detecting an attack, there is a need for a defense protocol to react to the attack. In this project, our objectives are studying and proposing mechanisms for devices in vehicular networks (vehicles installed communication units) to react to jamming attacking incidents in order to recover the communication in the networks.
– DataLake (Research on developing an application framework to support the collection, storage and exploitation of interdisciplinary scientific and technological data, from 2020-2022): The research team has successfully built a software system capable of storing and managing large-scale interdisciplinary science and technology data based on microservice architecture. The system is currently being tested on the server infrastructure of the ICTLab at USTH. With the current version, the system is capable of storing data in many different formats (demo in lung medical image format and digital format of Red River water quality data), allowing storage capacity of 06 TB (expandable to arbitrary quantity), allowing distribution on 04 nodes (expandable to arbitrary quantity), with high availability even when 01 hard drive fails. The software system also provides data manipulation services including collecting, integrating, querying, extracting and sharing lung medical image data and data on Red River surface water quality.
– iMorph (Detection of geometric morphometric landmarks in 2D insect wing images, from 2020-2022): The field of morphometrics, or morphometry, is concerned with the analysis of form, which is defined by shape and size, of an object. Shape is defined as the set of geometric properties of an object that are invariant to position, orientation, and scale. In a population of specimens of interest, shape variability with size, i.e. allometry, is also considered. The main goal of morphometrics is to elucidate how shapes vary and their covariance with other variables such as diseases, environmental stresses, or development etc. Morphometrics is very important in biology because it allows quantitative descriptions of organisms, hence facilitating comparative studies by statistical analysis methods. Landmark-based geometric morphometrics uses a set of landmarks to describe shape. Landmark, a two- or three-dimensional point represented by locus coordinates, is described by a tightly defined set of rules that indicates the evolutionary significance for the organism in question. Given these defined rules, it is next necessary to identify the landmarks on each specimen. This task is normally done by an expert, however, it is time-consuming and error-prone. Therefore, in this project, we aim to automate this by processing object images and employing an image recognition & machine learning model to propose landmark candidates.
– GEMMES (2019-2022). Assessment of the impact of Climate Change in Vietnam. Work Package leader of the WP4 Regio about the assessment of the local impact of climate change.
– LittoKong (2019). Sensitization of managers to the prevention of water-related risks: modeling, simulation, and participation in the Mekong Delta. Project in collaboration with the University of Tours, the University of La Rochelle and the University of Can Tho (Vietnam).
– HoanKiemAir (2019). Toward a tangible and interactive interface for the simulation of traffic and pollution in the Hoan Kiem district (Hanoi, Vietnam). Project in collaboration with University Thuy Loi.
– LungCancerCare (A system to support doctors in Lung Cancer Diagnosis and Treatment, from 2018-2019): Lung cancer is one of the most serious and common types of cancer all over the world, both in number of new patients and in number of fatalities. There were a total 1.69 millions of fatalities because of lung cancer in 2015 only (source: World Health Organization). Lung cancer is the most common type of cancer for both genders in Vietnam. According to statistics from the Ministry of Health, Vietnam has 25200 new patients in 2017 and expected 30.000 new patents per year in 2020. The goal of this project is to study and develop efficient deep learning models in order to improve the accuracy and computational performance of lung cancer image analysis algorithms. Specific goals include: (1) Building an annotated image dataset for lung cancer analysis with case studies in Vietnamese hospitals; (2) Studying and proposing deep learning models for efficient detection and classification of lung tumors as benign and malignant; (3) Developing a decision support software to assist doctors in lung cancer detection.
– HAGEDL (Real time Hand Gesture Recognition based on Deep Learning): Hand gesture is one of the most natural and intuitive ways people use to communicate and express emotions. Adoption of hand gestures in Human Machine Interaction (HMI) promises to provide users with comfort and ease. We introduce in this paper a vision-based recognition system for static hand gestures. Our method utilizes YOLOv3, a state of the art recognition deep neural network; therefore, it does not rely on complex image processing pipelines like many traditional systems. Experiments show that our framework is capable of processing full hd videos in real time speed with high accuracy.
– ESCAPE is an ANR project that focuses on the simulation of urban area evacuation in case of catastrophe. The simulator is based on the gamma simulation platform and is based on agent-based modeling of the evacuation process: this allows to explore individually based strategy, including individual knowledge about evacuation plan, emotion during egress and individually based mobility model with several modes. The project focuses on three case studies: chemical risk with industry explosion in the center of the city of Rouen (France), flash flood risk at the valley of Authion (France) and Hoa Binh dam break in the region of Hanoi – Red River (Vietnam).
– Gen* is an open source java library that makes it possible to generate spatially explicit and socially connected synthetic populations using any survey and GIS data. The toolkit is also available as a Gama plugin and can be used through the Kepler workflow management tool. The aim is to provide computer scientists and non programmers the access to state of the art algorithms to solve synthetic population generation, explicit localization and network generation of synthetic entities. The library is under development but already provides several algorithms for each of this three part: gospl, spll and spin. See https://github.com/ANRGenstar/genstar for further information.
– AgroLD (The Agronomic Linked Data project): Recent advances in high-throughput technologies have resulted in tremendous increase in the amount of data in the agronomic domain. This data explosion in-conjunction with its heterogeneity presents a major challenge in adopting an integrative approach towards research. We are developing AgroLD, a knowledge system that exploits the Semantic Web technology and some of the relevant standard domain ontologies, to integrate information on rice species and in this way facilitating the formulation of new scientific hypotheses. The objective of this effort is to provide the community with a platform for domain specific knowledge, capable of answering complex biological questions. The current phase covers information on genes, proteins, ontology associations, homology predictions, metabolic pathways, plant traits, and germplasm, on the Arabidopsis and rice species.
