Improving the ability of autonomous vehicles to accurately identify and follow lanes in various contexts is crucial. This project aims to provide a novel framework for optimizing a self-driving vehicle that can detect lanes and steer accordingly. A virtual sandbox environment was developed in Unity3D that provides a semi-automated procedural road and driving generation framework for a variety of road scenarios. Four types of segments replicate actual driving situations by directing the car using strategically positioned waypoints. A training dataset thus generated was used to train a behavioral driving model that employs a convolutional neural network to detect the lane and ensure that the car steers autonomously to remain within lane boundaries. The model was evaluated on real-world driving footage from Comma.ai, exhibiting an autonomy of 77% in low challenge road conditions and of 66% on roads with sharper turns.

Petrol is considered the most common fire accelerant. However, the identification and classification of petrol sources through the years has proven to be a challenging field in the investigation of fire debris analysis. This research explored the possibility of identifying petrol sources by high-field NMR methods accompanied by ML (machine learning). The automated identification and classification of petrol brands were achieved for first time based on the ML classification model developed in this research. A hierarchical classification model was constructed using local classifiers to categorize neat or weathered petrol into its sources.

Petrol is considered the most common fire accelerant. However, the identification and classification of petrol sources through the years has been proven to be a challenging field in the investigation of fire debris analysis. This research explored the possibility of identifying petrol sources by high field NMR methods accompanied by ML (Machine Learning). The automated identification and classification of petrol brands were achieved for first time based on the ML clas-sification model developed in this research. A hierarchical classification model was constructed using local classifiers to categorize neat or weathered petrol into its sources.

Accurate detection of rare-cell populations, such as circulating tumor cells (CTCs), is essential for early cancer diagnosis and personalized treatment. Imaging flow cytometry (IFC) enables high-throughput, single-cell imaging, but remains limited by operator-intensive analysis and challenges in sensitivity, scalability, and interpretability. We introduce attention-augmented image-only and multimodal convolutional neural network (CNN) architectures together with a streamlined preprocessing pipeline (multi-stage quality control, cropping, resizing, scaling, and channel stacking) that improve feature extraction and deliver robust, high-accuracy rare-cell detection in IFC. Beyond standard metrics, we evaluate models with synthetic spike-in simulations to mimic clinical class imbalance and use Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize decision-relevant regions. Experimental results show that multimodal architectures—fusing image embeddings with biologically meaningful metadata—generalize effectively to noisy, heterogeneous data and outperform conventional CNNs, achieving higher detection accuracy with fewer false positives while remaining computationally efficient. In particular, the attention-augmented multimodal model with aspect ratio and nuclear intensity (AttMM-2F) maintains high single-cell PC-3 recall (≈ 0.95) while keeping false positives low (≈ 1.6 per 5,000 peripheral blood lymphocytes (PBLs)) at a practical threshold. Notably, whereas many IFC phenotyping pipelines rely on additional biomarker channels, AttMM-2F achieves robust phenotyping using only brightfield and nuclear (DRAQ5, a DNA-binding nuclear stain) channels, avoiding additional immunostains and their associated time and reagent costs. Overall, the framework is interpretable and scalable, and we evaluate it with deployment-oriented protocols (threshold selection under rare-event prevalence and robustness stress tests). External multi-site and patient-derived validation remains a necessary next step before clinical deployment.