Browsing by Author "Rashmi Adyapady, R."

Now showing 1 - 5 of 5

An Xception Model with Residual Attention Mechanism for Facial Occlusion Detection
(Institute of Electrical and Electronics Engineers Inc., 2023) Rashmi Adyapady, R.; Annappa, B.
Occlusions occur due to the presence of obstacles. It poses difficulty in localizing and detecting the facial region, resulting in substantial intra-expression variability caused by noise and outliers. Facial occlusions are one of the most common issues that exist in real-world images. Solving such issues is essential for improving face recognition. The main aim of this work is to detect the occluded face. This work proposes a modified Xception network along with a residual attention mechanism to detect occluded parts of the facial region. The recognition accuracy obtained with the proposed Xception network with residual attention (Xcep-RA) mechanism is 99.97%, 99.85%, and 98.95% using Webface-OCC, Labeled Faces in the Wild (LFW), and Real-World Masked Face Dataset (RMFD) datasets. Extensive experiments using Xcep-RA significantly achieved competitive results compared to state-of-the-art methods on Webface-OCC, LFW, and RMFD datasets. Â© 2023 IEEE.
Cross-Database Facial Expression Recognition using CNN with Attention Mechanism
(Institute of Electrical and Electronics Engineers Inc., 2023) Chandra, J.; Annappa, B.; Rashmi Adyapady, R.
Facial expression is one of the most effective and universal ways to express emotions and intentions. It reflects what a person is thinking or experiencing. Thus, the expression recognition is one of the key aspects of understanding non-verbal communication and interpreting emotions in social interactions. Some emotions are very confusing, and separating the features between them becomes difficult because they share the same feature space. For example, the distinction between fear, anger, and disgust is confusing. This work tried to improve the model's class-wise performance to detect each class correctly. A distinct combination of deep-learning models is used to calculate the performance of the model, such as ResNet, XceptionNet, DenseNet, etc. The datasets like Real-world Affective Faces Database (RAF-DB), Japanese Female Facial Expression (JAFFE) & Facial Expression Recognition 2013 Plus (FER+) are used to evaluate the model's performance. The proposed model achieved better results and overcame the previous work's limitations. CDE's performance on RAF-DB and FER+ evaluations was significantly better than the current SOTA methods, with an increase in accuracy of 5.18% and 3.98%, respectively. Â© 2023 IEEE.
Learning Engagement Assessment in MOOC Scenario
(Institute of Electrical and Electronics Engineers Inc., 2022) Rashmi Adyapady, R.; Annappa, A.
Engagement recognition is essential for monitoring online learning for efficient learning outcomes. By monitoring the student's engagement, the teacher will acquire timely feedback, diminish the dropout rates, and overcome educational problems. A novel Facial Engagement Analysis-Network (FEA-Net) is proposed for learning engagement assessment in Massive Open Online Courses (MOOC) scenarios. In a MOOC setting, the combination of spatio-temporal and OpenFace features fed into FEA-Net proved effective for classifying engagement levels. The proposed FEA-Net built using Depthwise Separable Convolution layer helped improve the system's performance by reducing the model complexity. The experiment results showed an improvement of 1.01% in terms of accuracy on the Dataset for Affective States in E-learning Environments (DAiSEE) dataset. Â© 2022 IEEE.
Partial Convolution U-Net for Inpainting Distorted Images
(Institute of Electrical and Electronics Engineers Inc., 2024) Rashmi Adyapady, R.; Annappa, B.; Sagar, P.
Image inpainting is a domain in which researchers have shown considerable interest, and when it comes to deep learning techniques, realistic problems become interesting and challenging. In image inpainting, a corrupted facial image with missing holes or significant holes can be restored and compared to the original image to see if it is real or fake. In addition to fixing the texture of the image and getting the image's high-level abstract properties, it may also recover semantic images such as human faces. In the field of image-inpainting models, the Attention model with features learned through semantic approaches and progressive networks has become particularly popular. The proposed model introduces (i) Attention blocks in each decoder layer of U-Net architecture and (ii) a hybrid loss function leveraging both Mean Square Error (MSE) and Mean Absolute Error (MAE). The proposed Attention-based U-Net showed remarkable performance with SSIM and PSNR by 0.1067 and 13.63, respectively, compared to the previous approaches. Â© 2024 IEEE.
YARS-IDS: A Novel IDS for Multi-Class Classification
(Institute of Electrical and Electronics Engineers Inc., 2023) Madwanna, Y.; Annappa, B.; Rashmi Adyapady, R.; Sneha, H.R.
An Intrusion Detection System (IDS) is a defence system that provides safety and security against different threats and attacks, acting as a wall of defence against attackers. As internet usage increases, IDSs are becoming an essential part of day-to-day life. Various Machine Learning (ML) and Deep Learning (DL) based IDS are available, and the domain of IDS is still evolving and growing. Here this paper proposes two DL-based IDSs, first is a combination of LuNet and Bidirectional LSTM (Bi-LSTM) and other is a combination of Temporal Convolutional Network (TCN), CNN and Bi-LSTM. Such IDS must be fed with an efficient number of samples to keep them updated and accurate. The first model has been trained and tested against two benchmark datasets, NSL-KDD and UNSW-NB15. The second model has been trained and tested against the NSL-KDD dataset. To overcome the insufficient number of samples, the models have used a technique called Synthetic Minority Oversampling Technique (SMOTE). These models provided better experimental outcomes than traditional ML-based approaches and many DL approaches. They have better results in classification accuracy and, detection rate. The classification accuracy of the first model for UNSW-NB15 and NSL-KDD is 82.19% and 98.87% respectively. The classification accuracy of the second model for NSL-KDD is 98.8%. Â© 2023 IEEE.