We are pleased to announce that our article has been published in Information Sciences (IF: 8.1). The manuscript was submitted on January 27, 2024 and accepted on May 13, 2024.

📌The PDF of the article can be downloaded freely: https://doi.org/10.1016/j.ins.2024.120751

Title: Few-shot class incremental learning via robust transformer approach

Abstract:
Few-Shot Class-Incremental Learning (FSCIL)presents an extension of the Class Incremental Learning (CIL)problem where a model is faced with the problem of data scarcity while addressing the Catastrophic Forgetting (CF)problem. This problem remains an open problem because all recent works are built upon the Convolutional Neural Networks (CNNs)performing sub-optimally compared to the transformer approaches. Our paper presents Robust Transformer Approach (ROBUSTA)built upon the Compact Convolutional Transformer (CCT). The issue of overfitting due to few samples is overcome with the notion of the stochastic classifier, where the classifier’s weights are sampled from a distribution with mean and variance vectors, thus increasing the likelihood of correct classifications, and the batch-norm layer to stabilize the training process. The issue of CFis dealt with the idea of delta parameters, small task-specific trainable parameters while keeping the backbone networks frozen. A non-parametric approach is developed to infer the delta parameters for the model’s predictions. The prototype rectification approach is applied to avoid biased prototype calculations due to the issue of data scarcity. The advantage of ROBUSTAis demonstrated through a series of experiments in the benchmark problems where it is capable of outperforming prior arts with big margins without any data augmentation protocols.

We are pleased to announce that our article has been accepted in IEEE Transactions on Human-Machine Systems (IEEE THMS, IF: 3.5) after a lengthy peer-review process. The manuscript was submitted on February 3rd, 2021 and accepted on June 8th, 2024.

We would like to send our gratitude to all parties supporting the publication of the manuscript including IEEE THMS editorial team members, all reviewers, Universitas Gadjah Mada (UGM) through their APC Token IEEE program, Universitas Islam Indonesia for supporting the overlength page excess fee, and Sciencemind Lab for their assistance and proofreading effort during the final phase of publication.

📌 The PDF of the paper is free to download: https://doi.org/10.1109/THMS.2024.3413781

Title: Robust Object Selection in Spontaneous Gaze-Controlled Application Using Exponential Moving Average and Hidden Markov Model

Abstract:
The human gaze is a promising input modality for interactive applications due to its advantages: giving benefits to motion-impaired people while providing faster, intuitive, and easy interaction. The most common form of gaze interaction is object selection. During the last decade, gaze gestures and smooth pursuit-based interaction have been emerging techniques for spontaneous object selection in various gaze-controlled applications. Unfortunately, the challenge of spontaneous interaction demands no prior gaze-to-screen calibration, which leads to inaccurate object selection. To overcome the accuracy issue, this article proposes a novel method for spontaneous gaze interaction based on Pearson product-moment correlation as a measure of similarity, an exponential moving average filter for signal denoising, and a hidden Markov model to perform eye movement classification. Based on experimental results, our approach yielded the best object selection accuracy and success time of 89.60±10.59%
and 4364±235.86 ms, respectively. Our results imply that spontaneous interaction for gaze-controlled applications is possible with careful consideration of the underlying techniques to handle noisy data generated by the eye tracker. Furthermore, the proposed method is promising for future development of interactive touchless display systems that comply with the health protocols of the World Health Organization during the COVID-19 pandemic.

• Gravatar