The deep learning-based framework DeepExtractor reconstructs power excesses, including both signals and glitches, from Gravitational Wave (GW) data. The package includes scripts for generating training data, training models, and for evaluating performance, and contains a tutorial notebook for using DeepExtractor to reconstruct real glitches from O3 data at the LIGO detectors. DeepExtractor facilitates easy reconstruction of glitches and visualization of the results.