Thermography is an imaging technique harmless, low-cost and holding promise in detecting breast cancer at early stages. However, only with modern advances in infrared cameras and computational analysis, as machine learning based detection systems, it has achieved a sufficient accuracy to fulfill the role of adjunctive tool for breast abnormality detection. The automated segmentation of breast thermograms is crucial in improving the diagnostic utility of infrared thermography, as it precise localize regions of medical interest. This project propose a modular framework to address multiclass segmentation, exploiting open-source codebase deep-learning networks, in a smooth pipeline. This project aims to automatic detect and segment left and right breast, left and right nipples.