Publications

Below are summaries of my key publications:

Real-time tool detection in smart manufacturing using You-Only-Look-Once (YOLO) v5.

In this paper, we explore the use of the You-Only-Look-Once (YOLO) v5 object detection technique in real-time tool identification within a manufacturing setting. A dataset of over 3,000 tool images was collected and annotated to train the neural network. The model was optimized for high accuracy and generalization through a variety of techniques, such as image augmentation and stochastic gradient descent. The resulting model achieved a mean average accuracy of 98.3%, underscoring its effectiveness in real-time tool detection.

Real-time human-computer interaction using eye gazes

This research paper presents a real-time HCI system that utilizes eye gaze as a unique channel to recognize human intention. The system captures and tracks eyes using the Dlib 68-point landmark detector and employs an instance segmentation model using Mask R-CNN. The resulting system achieved an eye gaze recognition accuracy of over 99% and an instance segmentation model accuracy of 99%, demonstrating its feasibility and robustness in real-time operation.

Fine-grained activity classification in assembly based on multi-visual modalities

This study introduces a two-stage network for the recognition and prediction of fine-grained assembly activities in a manufacturing setting. The model combines scene-level features captured through RGB and hand skeleton frames with temporal-level features learned through RNN layers. The paper showcases robust performance, outperforming state-of-the-art models with an accuracy of over 99% in specific activity recognition and over 97% in activity prediction based on partial video observations.

Adaptive Robust Control for Trajectory Tracking of Autonomous underwater Vehicles on Horizontal Plane

This paper addresses the trajectory tracking problem of autonomous underwater vehicles in the horizontal plane. Adaptive sliding mode control is employed in order to achieve a robust behavior against some uncertainty and ocean current disturbances. The proposed approach is based upon a dual layer adaptive law, which is independent from the knowledge of disturbance boundary limit and its derivative. This approach tends to play a significant role in reducing the chattering effect that is prevalent in the conventional sliding mode controllers. 

For a comprehensive list of my publications, please visit my Google Scholar profile.