E. coli agar plate image dataset and fine-tuned machine-learning models

This deposition provides the imaging data, ground-truth annotations, and trained machine-learning model used to count Escherichia coli colony-forming units (CFUs) on agar plates with the Mycol application (see related works). It comprises: Plate images - 65 smartphone images of E. coli CFUs on LB agar (TIFF, 512 × 512 pixels, RGB). Plates were photographed on a white-light transilluminator using an iPhone 16 Pro or a Google Pixel 6a; colonies are dense, frequently occur near the plate edge, and some plates carry sticker or marker labels. Segmentation masks - 65 corresponding masks defining individual colony boundaries, one per image (TIFF), covering 12,150 annotated colonies. Trained model - a Cellpose cyto2 model (.pt) fine-tuned in Mycol on 13 of these images. The inference settings used are given in cellpose_inference_hyperparameters.csv. See Mycol GitHub notebook for image preprocessing requirements.