http://210.212.227.212:8080/xmlui/handle/123456789/510 2026-05-17T00:05:30Z http://210.212.227.212:8080/xmlui/handle/123456789/520 CAN INTRUSION DETECTION SYSTEM USING GAN Sreelakshmi, S Baiju; Adarsh, S Throughout the history of automobiles, advancements have been made to improve the safety and comfort of driving. One of the latest developments involves replacing the wiring between electronic control units (ECUs) with a networking standard called a Controller Area Network (CAN). While CAN has proven to be an effective communication protocol, it lacks security features that could not prevent malicious activities on the network. Therefore, there is a need for an Intrusion detection system (IDS) that can monitor CAN network traffic and identify any suspicious behavior. This work proposes an IDS for in-vehicle network communication that can detect both known and unknown malicious activities using deep learning techniques. The proposed IDS is based on Generative Adversarial Networks (GANs) which offers several novel features compared to traditional IDS techniques. The proposed IDS GAN model is evaluated using the Real ORNL Automotive Dynamometer (ROAD) CAN Intrusion Dataset, which contains many network traffic samples. The results shows that the model achieves high accuracy of 99%. Also had done a comparison with different enhanced CNN models to detect known attacks. It is evident from the above experiments that the model based on GANs can effectively detect network attacks and has the potential to be applied in real-world scenarios to enhance network security. 2023-05-01T00:00:00Z http://210.212.227.212:8080/xmlui/handle/123456789/519 DRIVER MANEUVER PREDICTION VIA FEATURE FUSION USING DEEP LEARNING Sarath, Mohan; Adarsh, S The eighth most common cause of death and the top cause of death for people aged 5-29 are traffic accidents. Automated vehicles and driver-assistance systems have emerged as a promising alternative to lower the number of fatalities in traffic accidents and provide a safer and more effective transportation system. Accurate driver maneuver prediction in a dynamic traffic scene remains a difficult topic due to its complexity, despite the considerable attention of researchers and industry. Driver maneuver prediction is a technique used by advanced driving assistance systems (ADAS) systems to give the driver early warnings and help. For instance, the system can inform the user if the driver is about to make a lane change without indicating. This project deals with accurately predicting the maneuver of the driver with the help of a deep learning model that utilizes feature fusion from various data. The dataset is simulated from the Car Learning to Act simulator (CARLA) from which driver face video, road video, and data from 12 sensors of the car are logged at 20 Hertz. The model incorporates U-shaped encoder-decoder network architecture (UNET) trained with the DRI(EYE)VE dataset to gauge the points of attention of the driver which is then used to extract the points of interest from the road video. A face landmark model is also incorporated by the model to retrieve essential features from the driver face video. These three features are fused and are fed into an long short term memory (LSTM) model for contextual learning and finally, the maneuver at a certain time ahead is classified. Experimental results have shown that the feature fusion model obtained an accuracy of 80.19%, an overall precision of 87.93%, an overall recall of 87.95%, and an F1 score of 87.80%. 2023-05-01T00:00:00Z http://210.212.227.212:8080/xmlui/handle/123456789/518 INTELLIGENT POTHOLE DETECTION AND ASSESSMENT SYSTEM: AREA AND DEPTH ESTIMATION Rufus Rubin, Oscar Fernandez; Chinnu, Jacob Economic and social prosperity relies heavily on well-developed and maintained highways, but lack of funding and resources makes highway maintenance difficult. Potholes and early road deterioration pose significant risks for automobile accidents. Therefore, it is essential to identify and repair potholes promptly to maintain dependable and safe road infrastructure. To address this issue, the YOLO(You Only Look Once) detection algorithm is used to detect potholes in this work. The proposed approach includes a Single Pothole Detector based on the YOLO-v1 Algorithm and a Multiple Pothole Detector based on the YOLO-v3 and YOLO-v5 Algorithms. Various backbone architectures are utilized to identify potholes in the Single Pothole Detector experiment. In addition, attention mechanisms are integrated into these backbone architectures to improve performance in pothole detection, and computer vision methods are employed to estimate the area of potholes. The models are trained and validated using a Custom dataset comprising the MakeML pothole dataset, a custom dataset from Kaggle, and two real-time footage of Kerala roadways. Additionally, Depth estimation is being carried out using Encoder-Decoder architecture. The Pothole Detectors’ performance is evaluated using mean Average Precision (mAP) and Average Precision(AP) metrics to compare them with other models. The results show that the proposed Pothole Detectors perform better than other models, highlighting their potential in identifying and addressing potholes promptly to maintain safe and dependable road infrastructure. 2023-05-01T00:00:00Z http://210.212.227.212:8080/xmlui/handle/123456789/517 WAVELET BASED CNN FOR DIAGNOSIS OF LUNG DISEASES USING CHEST X-RAYS Jumana, R; Chinnu, Jacob Lung diseases are widespread worldwide, including conditions such as lung nodules, pneu monia, asthma, tuberculosis, fibrosis, etc. It is crucial to diagnose lung disease promptly. In this study, wavelet-based Convolutional Neural Networks have been employed to detect and differentiate various types of lung diseases through analyzing chest X-ray images. Existing CNN architectures have been used previously to classify healthy and affected condition chest X-rays. However, these networks process the image in a single resolution and may lose poten tial features present in other resolutions of the input image. Over this normal CNN, Wavelets are utilized to decompose the image into different spatial resolutions based on high pass and low pass frequency components and extract valuable features from the affected portion of lung X-ray images efficiently. A CNN model of wavelet is employed to find relevant features from the X-ray images, and SVM classifier is incorporated to classify different lung diseases from the extracted features. The proposed framework is tested on three publicly available datasets and the method achieved an average accuracy of 100% using the first dataset (NIH Chest X-ray), 100% accuracy on the second dataset (LUNG DISEASE), and 96% accuracy on the third dataset (JSRT). Overall, the proposed approach outperformed existing works and demonstrated its effectiveness in identifying multiple lung conditions, including nodules, COVID-19, and other ailments, making it a versatile tool 2023-07-01T00:00:00Z