Publications

2024

1. Journal

   A Comprehensive Analysis of Cognitive CAPTCHAs through Eye Tracking

   Nghia Dinh, Lidia Ogiela, Kiet Tran-Trung, Tuan Le-Viet and Vinh Truong Hoang

   IEEE Access, 2024

   Abs PDF

   CAPTCHA (Completely Automated Public Turing Test to Tell Computers and Humans Apart) has long been employed to combat automated bots. It accomplishes this by utilizing distortion techniques and cognitive characteristics. When it comes to countering security attacks, cognitive CAPTCHA methods have proven to be more effective than other approaches. The advancement of eyetracking technology has greatly improved human-computer interaction (HCI), enabling users to engage with computers without physical contact. This technology is widely used for studying attention, cognitive processes, and performance. In this specific research, we conducted eye-tracking experiments on participants to investigate how their visual behavior changes as the complexity of cognitive CAPTCHAs varies. By analyzing the distribution of eye gaze on each level of CAPTCHA, we can assess users' visual behavior based on eye movement performance and process metrics. The data collected is then employed in Machine Learning (ML) algorithms to categorize and examine the relative importance of these factors in predicting performance. This study highlights the potential to enhance any cognitive CAPTCHA model by gaining insights into the underlying cognitive processes.

2023

1. Journal

   MoGuP: Motion-guided Prediction for Video Anomaly Detection

   Viet-Tuan Le and Yong-Guk Kim

   Under submission

   Abs PDF Website Code

   Video anomaly detection aims to identify sporadic abnormal events among abundant normal ones in surveillance videos. For this purpose, many existing methods leverage U-Net architecture to predict or reconstruct frames, typically extracting spatial features within a sequence of previous consecutive frames. Yet, several approaches utilize motion information by introducing a two-stream network where the spatial and temporal streams flow independently, fusing the final outputs from these streams to obtain the anomaly score. Given a target optical flow and a sequence of previous frames, it would be desirable to design a framework that predicts the target frame with the intended motion corresponding to the provided optical flow. We present a novel motion-guided prediction framework that uses a proposed motion module to guide the prediction process for generating the output, that is close to its ground truth. Extensive evaluation, including ablation studies, suggests that the proposed approach outperforms the state-of-the-art on three standard benchmark datasets, i.e. UCSD Ped2, CUHK Avenue, and ShanghaiTech. This results suggest that it has the potential for real applications. The code and model are available at https://moguprediction.github.io.

2. Journal

   HSTforU: Anomaly Detection in Aerial and Ground-based Videos with Hierarchical Spatio-Temporal Transformer for U-net

   Viet-Tuan Le, Hulin Jin, and Yong-Guk Kim

   Under review

   Abs PDF Website Code

   Anomaly detection is to identify abnormal events against normal ones within surveillance videos mainly collected in ground-based settings. Recently, the need for drone data processing is rapidly growing as drones find new applications. However, as most aerial videos collected by flying drones contain moving backgrounds and others, it is necessary to deal with their spatio-temporal features in detecting anomalies. This study presents a transformer-based video anomaly detection method whereby we investigate a challenging aerial dataset and three benchmark ground-based datasets. The encoder of our U-net has a four-stage pyramid transformer structure, and each stage has a link to a corresponding spatio-temporal transformer stage. Then, this transformer produces hierarchical feature maps that are conveyed to the decoder as skip connections. Extensive evaluations including several ablation studies suggest that this network outperforms the state-of-the-art on Drone-anomaly dataset and three benchmark datasets. We expect the proposed transformer for U-net can find diverse applications in the video processing area. Code and model are available at https://vt-le.github.io/HSTforU/.

3. ICIT

   Detection of Kidney Stone Based on Super Resolution Techniques and YOLOv7 Under Limited Training Samples

   Minh Tai Pham Nguyen, Viet Tuan Le, Huu Thanh Duong, and Vinh Truong Hoang

   Intelligence of Things: Technologies and Applications, 2023

   Abs PDF

   In machine vision, detecting and locating object have always been challenging tasks mostly in medical imaging due to the lacking of data and unique feature of its image. Several studies applied CNN-based models for detecting task but the methods were complex or impracticable to real time tasks while others used YOLO models which were lighter but poor performance. In this study, we propose Super Resolution techniques included EDSR, FSRCNN, LapSRN, ESPCN, ESRGAN and GFPGAN with the proposed architecture of YOLOv7 for kidney stone detection on KUB X-ray image. As a result, our proposed YOLOv7 outperformed the YOLOv7 base version with sensitivity 87.6%, Precision 92.2%, F1 Score 89.8%, mAP50 91.2% and each of Super Resolution methods enabled the model precision and sensitivity to be improved considerably with highest precision reached 97.3% and sensitivity achieved 91.7% on upscaled images compared to the non-upscaled images. Consequently, the re-designed YOLOv7 and Super Resolution methods are proposed to address the problem of detection in diagnosis kidney stone disease.

4. ICIT

   Multi-stage Transformer for Video Anomaly Detection

   Viet Tuan Le, Khuong G. T. Diep, Tae-seok Kim, and Yong-guk Kim

   Annual Conference of KIPS, 2023

   Abs PDF

   Video anomaly detection aims to detect abnormal events. Motivated by the power of transformers recently shown in vision tasks, we propose a novel transformer-based network for video anomaly detection. To capture long-range information in video, we employ a multi-scale transformer as an encoder. A convolutional decoder is utilized to predict the future frame from the extracted multi-scale feature maps. The proposed method is evaluated on three benchmark datasets: USCD Ped2, CUHK Avenue, and ShanghaiTech. The results show that the proposed method achieves better performance compared to recent methods.

5. ICIT

   Novel Reward Function for Autonomous Drone Navigating in Indoor Environment

   Khuong G. T. Diep, Viet Tuan Le, Tae-seok Kim, Anh H. Vo, and Yong-guk Kim

   Annual Conference of KIPS, 2023

   Abs PDF

   Unmanned aerial vehicles are gaining in popularity with the development of science and technology, and are being used for a wide range of purposes, including surveillance, rescue, delivery of goods, and data collection. In particular, the ability to avoid obstacles during navigation without human oversight is one of the essential capabilities that a drone must possess. Many works currently have solved this problem by implementing deep reinforcement learning (DRL) model. The essential core of a DRL model is reward function. Therefore, this paper proposes a new reward function with appropriate action space and employs dueling double deep Q-Networks to train a drone to navigate in indoor environment without collision.

2022

1. Journal

   Attention-based Residual Autoencoder for Video Anomaly Detection

   Viet-Tuan Le and Yong-Guk Kim

   Applied Intelligence, 2022

   Abs PDF Website Code

   Automatic anomaly detection is a crucial task in video surveillance system intensively used for public safety and others. The present system adopts a spatial branch and a temporal branch in a unified network that exploits both spatial and temporal information effectively. The network has a residual autoencoder architecture, consisting of a deep convolutional neural network-based encoder and a multi-stage channel attention-based decoder, trained in an unsupervised manner. The temporal shift method is used for exploiting the temporal feature, whereas the contextual dependency is extracted by channel attention modules. System performance is evaluated using three standard benchmark datasets. Result suggests that our network outperforms the state-of-the-art methods, achieving 97.4% for UCSD Ped2, 86.7% for CUHK Avenue, and 73.6% for ShanghaiTech dataset in term of Area Under Curve, respectively.

6. Journal

   A comprehensive review of recent deep learning techniques for human activity recognition

   Viet-Tuan Le, Kiet Tran-Trung, and Vinh Truong Hoang

   Computational Intelligence and Neuroscience, 2022

   Abs PDF

   Human action recognition is an important field in computer vision that has attracted remarkable attention from researchers. This survey aims to provide a comprehensive overview of recent human action recognition approaches based on deep learning using RGB video data. Our work divides recent deep learning-based methods into five different categories to provide a comprehensive overview for researchers who are interested in this field of computer vision. Moreover, a pure-transformer architecture (convolution-free) has outperformed its convolutional counterparts in many fields of computer vision recently. Our work also provides recent convolution-free-based methods which replaced convolution networks with the transformer networks that achieved state-of-the-art results on many human action recognition datasets. Firstly, we discuss proposed methods based on a 2D convolutional neural network. Then, methods based on a recurrent neural network which is used to capture motion information are discussed. 3D convolutional neural network-based methods are used in many recent approaches to capture both spatial and temporal information in videos. However, with long action videos, multistream approaches with different streams to encode different features are reviewed. We also compare the performance of recently proposed methods on four popular benchmark datasets. We review 26 benchmark datasets for human action recognition. Some potential research directions are discussed to conclude this survey.

2019

1. ICSPS Oral

   Local binary pattern based on image gradient for bark image classification

   Tuan Le-Viet and Vinh Truong Hoang

   Tenth International Conference on Signal Processing Systems (ICSPS), Singapore, 2019

   Abs PDF

   In this work, we present a discriminative and effective local texture descriptor for bark image classification. The proposed descriptor is based on three factors, namely, pixel, magnitude and direction value. Unlike most other descriptors based on original local binary pattern, the proposed descriptor is conducted the changing of local texture of bark image. The performance of the proposed descriptor is evaluated on three benchmark datasets. The experimental results show that our approach is highly effective.

2012

1. ICCAIS

   Human action recognition on simple and complex background in video

   Tuan Le-Viet and Ngoc Quoc Ly

   International Conference on Control, Automation and Information Sciences, 2012

   Abs PDF

   In this paper, we propose a new method for human action recognition on simple and complex background. There are two approaches: (a) for homogeneous and static background, we propose a combination of HOG3D and HOF to encode the appearance and motion information of human. (b) with realistic video data (including changes of viewpoint, scale, and lighting conditions, partial occlusion of humans and objects, cluttered backgrounds), we exploit the co-occurrence between scene and actions by combining HOG3D and Opponent SIFT to encode action (how) and scene (where) information. The first combination is applied in surveillance videos or videos in which camera placement and parameters are fixed and known. The second one is applied to remained video in which background is not static and cluttered. The experimental results have shown the efficiency and effectiveness of our method.