The Breakthrough Listen Initiative collects data using the Green Bank Telescope and aims to search for extra-terrestrial intelligence (ETI) signals. To search for these signals in hundreds of gigabytes of raw observations is computationally extensive. Thus, it is beneficial to use Gaussian filtering and object detection models for the purpose of data cleaning and identifying potential ETI signals. Based on observations, we assume that background noise is of Gaussian distribution, and we can detect any excess or deviation from it by performing a normal test on portions of the data collected. Thus, we choose a fitting p-value threshold to pick out these signals of interest and save them as samples for further processing. This reduces the size of the dataset by 80%. Next, we use a state-of-the-art object detection model, a Mask Region-based Convolutional Neural Network (Mask-RCNN) pre-trained on ImageNet, to draw out bounding boxes that localize signals on these samples. Specifically, the model predicts there is a high probability that the regions defined by the bounding boxes contain objects and in this case the important features of the data. Using Gaussian filtering and object detection, we are able to remove background noise and extract out key features of the data from large amounts of raw observational data. Any further processing and/or searches for ETI signals can be performed on this cleaned and more compact dataset. This work is conducted by the BL@Scale Team under the auspices of Breakthrough Listen at UC Berkeley.