Dr. Irfan Hussain's Lab- Physical and Cognitive Intelligence Robotics Group

🚀 Excited to announce the publication of our latest research in IEEE Access! Our study, titled “Unified Synergistic Deep Learning Framework for Multimodal 2-D and 3-D Radiographic Data Analysis: Model Development and Validation”, presents a breakthrough in medical imaging analysis.

💡 Our approach integrates 2-D and 3-D radiographic data using a unified synergistic deep learning model, enhancing diagnostic accuracy in clinical radiology. We leveraged multilevel features with a lightweight Vision Transformer and Multilevel-Multilayer Perceptron heads, achieving significant results with:

- Accuracy: 96.67%
- F1-Score: 96.98%
- True Positive Rate: 96.75%
- True Negative Rate: 97.02%

📊 Our method outperformed existing solutions on public radiographic datasets, showcasing its potential for automating lung infection detection through advanced multimodal imaging.

🔗 Read the full paper here: https://ieeexplore.ieee.org/abstract/document/10737310

🙏 Grateful for the collaboration with my fellow-authors: Dr. Muhammad Zubair, Prof. Lakmal Seneviratne, Prof. Naoufel Werghi, and Dr. Irfan Hussain

We had the distinct pleasure of organizing the technical visits for IROS 2024 at Khalifa University, where we have had the opportunity to showcase some of our key research activities. These visits have provided a platform to engage with fellow researchers and industry leaders, highlighting the innovative work taking place in the following labs:
🔹 Manipulation Lab – Showcasing advancements in robotic manipulation and automation.
🔹 Agritech Lab – Exploring cutting-edge technologies aimed at revolutionizing agriculture.
🔹 Marine Lab – Focusing on our pioneering research in marine robotics and ocean exploration.
It has been rewarding to see the interest and enthusiasm from our international visitors and to engage in meaningful discussions about potential collaborations. We extend our sincere thanks to all who have contributed to the success of these visits.

🎉 Excited to share our latest research presented at The 14th IEEE International Conference on Signal Processing, Communications, and Computing (ICSPCC 2024)! 🎉

📄 Paper Title: Enhanced Gesture Recognition through Graph-Based Multimodal Fusion

👨‍💻 Authors:

Mobeen Ur Rehman (KU Center for Autonomous Robotic Systems, Khalifa University)

Talha Ilyas (Monash Medical AI Group, Monash University)

Lakmal Seneviratne (KU Center for Autonomous Robotic Systems, Khalifa University)

Irfan Hussain (KU Center for Autonomous Robotic Systems, Khalifa University)

🔍 Abstract: We introduce an advanced framework for recognizing hand gestures from a first-person view, leveraging multimodal data (optical flow, pose, depth, and RGB video). The framework employs a cross-attention-based adaptive graph convolutional network and relational graph interactions for modality fusion. This approach captures nuanced hand movements and facilitates the fusion of heterogeneous data types, significantly enhancing classification accuracy and robustness.

🚀 Key Highlights:

Achieved an impressive 98.48% accuracy on a public benchmark dataset.

Maintains strong performance (93.48% accuracy) even with a single modality, showcasing its potential for real-world applications.

This breakthrough sets a new standard in hand gesture recognition, paving the way for future advancements in multimodal data fusion.

Special thanks to Khalifa University of Science and Technology (Award No. CIRA-2021-085, FSU-2021-019, RC1-2018-KUCARS) and KUCARS for their support and facilities.

We are happy to share that our manuscript titled “Deep learning approach for detecting tomato flowers and buds in greenhouses on 3P2R gantry robot”, supervised by Dr. Irfan Hussain, was published in Nature Scientific Reports Journal (Top 8% as per Scopus)!

In this work, we developed a methodology for labeling, training, and detecting tomato flowers tailored for robotic pollination using transfer learning with YOLOv5 and YOLOv8 models. The performance of both models was evaluated using an image dataset collected under varying lighting conditions. YOLOv8 achieved a higher mean Average Precision (mAP) of 92.6% and faster inference speed compared to YOLOv5, making it more suitable for real-time detection. The study also included deploying the YOLOv8 model on a 3P2R gantry robot using position-based visual servoing to approach detected flowers, which proved effective in both clustered and un-clustered environments. The research offers valuable insights into designing flower detection algorithms for greenhouse robotic systems.
The paper is available at: https://www.nature.com/articles/s41598-024-71013-1

Big congratulations to our team for this achievement: Dr. Asim Khan, Dr. Lakmal Seneviratne, and Dr. Irfan Hussain!

Thank you, Khalifa University and Khalifa University Center of Autonomous Robotics Systems for the continuous support.

We are happy to share that our manuscript titled “MuLA-GAN: Multi-Level Attention GAN for Enhanced Underwater Visibility”, supervised by Dr. Irfan Hussain, has been published in Ecological Informatics by Elsevier, with an impact factor of 4.87 and a top 10% percentile ranking.

In this work, we introduce MuLA-GAN, a novel approach that leverages Generative Adversarial Networks (GANs) integrated with Multi-Level Attention mechanisms to significantly enhance underwater image quality. Our innovative method effectively tackles common underwater imaging challenges, including color distortions, reduced contrast, and blurriness. By employing advanced techniques to prioritize and learn discriminative features, MuLA-GAN achieves significant performance improvements over existing methods. It delivers exceptional results with a Peak Signal-to-Noise Ratio (PSNR) of 25.59 and a Structural Similarity Index (SSIM) of 0.893 across diverse datasets, showcasing its robustness and effectiveness in improving underwater image clarity and detail.

The paper is available at: https://www.sciencedirect.com/science/article/pii/S1574954124001730

Big Congratulations to our team for this outstanding achievement. Ahsan Baidar Bakht, Zikai Jia, Muhayy Ud Din, Waseem Akram, Lyes Saad Saoud, Lakmal Seneviratne, Defu Lin, Shaoming He, and Dr. Irfan Hussain.

Thank you Khalifa University and Khalifa University Center of Autonomous Robotics Systems for the continuous support.

🚀 A new addition to our Manipulation Lab: Aubo i10!

We welcome a new robotic arm, Aubo i10, to our manipulation lab. This robotic arm is set to revolutionize our manipulation research and exploration capabilities. 🛠️🔬

With the Aubo i10, we aim to deploy it for our ongoing manipulation project related to robotic teleoperation, motion planning, and the application for smart farming with generative AI 🌱.

Stay tuned for updates on our upcoming expeditions and the incredible work we’ll be doing with this new tool.

Long-Range Vision-Based UAV-assisted Localization for Unmanned Surface Vehicles:

We present a novel method that utilizes an Unmanned Aerial Vehicle (UAV) to assist in localizing USVs in GNSS-restricted marine environments. In our approach, the UAV flies along the shoreline at a consistent altitude, continuously tracking and detecting the USV using a deep learning-based approach on camera images. Subsequently, triangulation techniques are applied to estimate the USV’s position relative to the UAV, utilizing geometric information and datalink range from the UAV. We propose adjusting the UAV’s camera angle based on the pixel error between the USV and the image center throughout the localization process to enhance accuracy. Additionally, visual measurements are integrated into an Extended Kalman Filter (EKF) for robust state estimation

This work is the result of a collaborative effort made by a talented team, including Waseem Akram, Siyuan Yang, Hailiang Kuang, Xiaoyu He, Muhayy Ud Din,Yihao D**g, Defu Lin, Lakmal Seneviratne, Shaoming He and Irfan Hussain.



https://www.researchgate.net/publication/383280040_Long-Range_Vision-Based_UAV-assisted_Localization_for_Unmanned_Surface_Vehicles