The recognition of microplastics (MPs) in environmental samples via FT-IR is challenging due to a plethora of factors can lead to significant variances in measured spectra. Conventional library search approaches compare the observed spectrum with spectra in reference libraries, which will lead to errors due the variance in spectra. Motivated to tackle this challenge, this study explores the feasibility of leveraging deep learning for automatic MP recognition via FT-IR spectroscopy. More specifically, a deep convolution neural network (CNN) architecture, referred to here as PlasticNet, is introduced for the purpose of automatic MP recognition. PlasticNet was trained on a large corpus of FT-IR spectra of different plastic types in order to learn discriminative spectral features characterizing each plastic type. Experimental results showed that PlasticNet was capable of recognizing between MPs in an effective way and at a faster speed compared with libary search.