Detecting radioactive materials in mixtures is challenging due to low concentration, environmental factors, sensor noise, and others. This article presents new results on nuclear material identification and mixing ratio estimation for mixtures of materials in which there are multiple isotopes present. Conventional and deep learning (DL)-based machine learning (ML) algorithms were compared. Both simulated and actual experimental data were used in the comparative studies. It was observed that some newly developed ML and DL methods have better performance than a conventional method based on the region of interest (ROI). Extensive experiments also demonstrated that trained models using low signal-to-background ratio (SBR) data can better generalize to other SBR conditions.