This is an annual report for 2024 of the signal classification project (Signalklassificering 23-25). During the year, research on modulation recognition using deep learning has been conducted, specifically with regards to the area of one-shot learning which concerns deep learning methods for classification of rare signals. A Relation Network has been studied, and parts of this research was the subject of a research paper presented at the IEEE conference MILCOM 2024. The Relation Network is an interesting architecture since it, besides from being able to perform one-shot learning, works as a comparator contrary to typical classifiers, allowing e.g. an operator to dynamically update the reference signal library without the need for retraining the network. The project also worked with estimating the wireless channel for use in deep learning models. This report will introduce the research conducted on these topics and provide some suggestions for future research.