Model Architecture

The model architecture will be based on Faster R-CNN, a state-of-the-art object detection CNN that uses a region proposal algorithm to hypothesize object locations \cite{Ren_2017}.  Depending on the size of the training data gathered, it may be augmented with synthetic plumes in order to avoid overfitting. For detection of temporal activity i.e. tracking of the plumes, the model will also build off of Region Convolutional 3D Network (R-C3D) which can be used to extract spatiotemporal features capturing activities, accurately localizing the start and end times of each plume \cite{saenko2017}.