There are multiple approaches available for identifying and quantifying the number of microplastic (MP) particles present in environmental samples. The application of the different techniques depends on multiple factors including instrument purchase and running costs, instrumental availability, ease of use, time for analysis, parameter(s) of interest and quality of the final data. Fluorescent staining and imaging of MP particles is a method that is cheap to implement, easy to conduct and which offers high potential for rapid screening of large numbers of samples. While there are clear and acknowledged limitations with respect to some of the other methods and analysis systems (e.g. lack of specificity in identification, false positives, sensitivity), one major issue preventing the wider acceptance of this approach as a screening technique is a lack of robust and reproducible automation in the processing of images containing fluorescent MP on filters. Manually analysing filters with stained MP is not only a tedious and time-consuming task, but it can be very inaccurate and suffers with reproducibility issues both within an individual lab and across labs. To improve accuracy and reproducibility, whilst reducing time and cost of analysis, we have developed an algorithm based on deep learning to automatically identify the stained MP. The algorithm consists of a custom Fully Convolutional Network (FCN) architecture for particle segmentation and classical image processing for acquiring metrics including individual particle area, length and width of each particle. This algorithm can process data from one filter (consisting of ten 50-megapixel images) in less than one minute, completely unattended.