The chaotic spatio-temporal electrical activity during life-threatening cardiac arrhythmias liketextlessbrtextgreaterventricular fbrillation is governed by the dynamics of vortex-like spiral or scroll waves. Thetextlessbrtextgreaterorganizing centers of these waves are called wave tips (2D) or flaments (3D) and they play a key roletextlessbrtextgreaterin understanding and controlling the complex and chaotic electrical dynamics. Therefore, in manytextlessbrtextgreaterexperimental and numerical setups it is required to detect the tips of the observed spiral waves. Mosttextlessbrtextgreaterof the currently used methods signifcantly sufer from the infuence of noise and are often adjustedtextlessbrtextgreaterto a specifc situation (e.g. a specifc numerical cardiac cell model). In this study, we use a specifc typetextlessbrtextgreaterof deep neural networks (UNet), for detecting spiral wave tips and show that this approach is robusttextlessbrtextgreateragainst the infuence of intermediate noise levels. Furthermore, we demonstrate that if the UNettextlessbrtextgreateris trained with a pool of numerical cell models, spiral wave tips in unknown cell models can also betextlessbrtextgreaterdetected reliably, suggesting that the UNet can in some sense learn the concept of spiral wave tips in atextlessbrtextgreatergeneral way, and thus could also be used in experimental situations in the future (ex-vivo, cell-culturetextlessbrtextgreateror optogenetic experiments).