To address this challenge, AlertCalifornia opted for a sophisticated solution that leveraged advanced machine learning techniques. The chosen approach involved the development of a custom Long Short-Term Memory (LSTM) model. Unlike traditional models that analyze individual images in isolation, the LSTM model was designed to capture temporal dependencies by learning from sequences of smoke bounding box data over time.
The workflow involved using PyTorch and PyTorch Lightning for model development, taking advantage of their flexibility and efficiency in implementing complex neural network architectures. Additionally, PySpark was utilized for handling large-scale data processing, ensuring that the model could be trained on extensive datasets effectively.
Amazon SageMaker, a fully managed service for building, training, and deploying machine learning models, played a crucial role in the training phase. The SageMaker service facilitated the seamless integration of the developed LSTM model into the production environment, offering a scalable and efficient platform for training and deploying machine learning models.
.svg.png){kind=logo}
