Road traffic noise as a pollutant is being investigated within an ongoing research project with Stockholm City. An important objective of this project is utilising smart-city infrastructure and pollutant propagation models for making reliable pollutant exposure assessments. To improve noise emission models, extensive noise measurements are being conducted on a test-bed in Södermalm, Stockholm. This data serves as input for advanced neural network architectures (e.g., CRNN under a student-teacher paradigm) for building a model that can detect and classify vehicles from road-side noise measurements. A properly trained model will enable reliably determining the strength of noise sources, and also differentiating vehicle-related noise from other sources. Training these models requires both large GPU memory, as well as, large storage for the extensive noise data used for training. These models have been successfully tested on limited a dataset, and can now be expanded to the entire dataset.