Abstract: The recognition of gas information based on deep learning methods faces the disadvantages of poor classification stability, low accuracy, and feature degradation. In this work, a residual ...