Rapid identification of bacterial species from patient samples is crucial for clinical decision-making. In severe infections, such as bloodstream infections, the early start of an effective treatment is directly associated with reduced mortality rates. Current rapid species identification methods, such as matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) or multiplex PCR, require specialized hardware and extensive technical support that prevents application in resource-limited settings. Here, we present a staining and imaging procedure for bacterial smears using fluorescent dyes directed against intracellular structures and cell wall components. Data on relevant features were extracted from segmented images and used to train a machine learning (ML) model for species classification. The method was tested on clinical isolates from 126 patients. For the seven most common bacteria, the classification performance, indicated by area under the receiver operating characteristic (ROC) curve, ranged from 0.8 (Klebsiella pneumoniae) to 1 (Pseudomonas aeruginosa). Species that were not part of the training dataset, were reliably classified as unknown species. These results hold promise for the identification of further species, particularly Enterobacterales, and clinical application.