TV Commercial Classification and Optimization of Visual Recognition Pipeline for News Videos (For Red Hen)

Red Hen has an existing visual feature based classification and labeling system for news videos, but TV commercials are not considered by the pipeline so far. I want to add a multimodal TV commercial classification system to the pipeline’s current capabilities. In addition to this, I will work on enriching the current state of the visual pipeline using semi supervised learning. The third major component of my project will be to optimize the performance of the deep learning pipeline in order to allow the system to efficiently scale and process the massive amount of data that the NewsScape dataset has.