The research describes an effective deep learning-based, data-centric approach for diagnosing autism spectrum disorder from facial images. To classify ASD and non-ASD subjects, this method requires training a convolutional neural network using the facial image dataset. As a part of the data-centric approach, this research applies pre-processing and synthesizing of the training dataset. The trained model is subsequently evaluated on an independent test set in order to assess the performance matrices of various data-centric approaches. The results reveal that the proposed method that simultaneously applies the pre-processing and augmentation approach on the training dataset outperforms the recent works, achieving excellent 98.9% prediction accuracy, sensitivity, and specificity while having 99.9% AUC. This work enhances the clarity and comprehensibility of the algorithm by integrating explainable AI techniques, providing clinicians with valuable and interpretable insights into the decision-making process of the ASD diagnosis model.
1 February 2024 in Bulletin of Electrical Engineering and Informatics
Road defect inspection is a crucial task in maintaining a good transportation infrastructure as road surface distress can impact user’s comfortability, reduce the lifetime of vehicles’ parts, and cause road casualties. In recent years, machine learning has been adapted widely in various fields, including object detection, thanks to its superior performance and the availability of high computing power which is generally needed for its model training. Many works have reported using machine-learning-based object detection algorithms to detect defects, such as cracks in buildings and roads. In this work, YOLOv5, YOLOv6 and YOLOv7 models have been implemented and trained using a custom dataset of road cracks and potholes and their performances have been evaluated and compared. Experiments on the dataset show that YOLOv7 has the highest performance with mAP@0.5 score of 79.0% and an inference speed of 0.47 m for 255 test images.
This paper addresses the development of a stereo vision-based obstacle avoidance system using MOOS-IvP for small and medium-sized Unmanned Surface Vehicles (USVs). Existing methods predominantly rely on optical sensors such as LiDAR and cameras to discern maritime obstacles within the short- to mid-range distances. Nonetheless, conventional cameras encounter challenges in water conditions that curtail their effectiveness in localizing obstacles and planning paths. Furthermore, LiDAR has limitations regarding angular resolution and identifying objectness due to data sparsity. To overcome these limitations, our proposed system leverages a stereo camera equipped with enhanced angular resolution to augment situational awareness. The system employs recursive estimation techniques to ascertain the position and dimensions of proximate obstacles, transmitting this information to the onboard control unit, where MOOS-IvP behaviour-based software produces navigation decisions. Through the real-time fusion of data obtained from the stereo vision system and navigational data, the system is able to achieve Enhance Situational Awareness (ESA) and facilitate well-informed navigation decisions. Developing a state-of-the-art maritime object detection technique, the system adeptly identifies obstacles and swiftly responds via a vision integration protocol. During field tests, our system proves the efficacy of the proposed ESA approach. This paper also presents a comprehensive analysis and discussion of the results derived from deploying the proposed system on the Suraya Surveyor USV platform across numerous scenarios featuring diverse obstacles. The results from these various scenarios demonstrate the system’s accurate obstacle detection capabilities under challenging conditions and highlight its significant potential for safe USV operations.