DeepBacs – Escherichia coli antibiotic phenotyping object detection dataset and YOLOv2 model
Creators
- 1. Institute of Physical and Theoretical Chemistry, Max-von-Laue Str. 7, Goethe-University Frankfurt, 60439 Frankfurt, Germany
Description
Training and test images of E. coli cells treated with different antibiotics for antibiotic phenotyping using YOLOv2 object detection.
Additional information can be found on this github wiki.
Example images show predictions of drug-treated E. coli cells.
Training and test dataset
Data type: Paired microscopy images (confocal fluorescence) and manual annotations
Microscopy data type: Confocal fluorescence images of fixed E. coli cells stained for membrane (Nile Red) and DNA (DAPI) paired with annotations in PASCAL VOC format
Microscope: Zeiss LSM710 confocal microscope with a Plan-Apo 63x oil objective (1.4 NA)
Cell type: Chemically fixed E. coli NO34 cells (MreB-sfGFPsw, kindly provided by Zemer Gitai) (untreated or drug-treated);
File format: .png (RGB)
Image size: 400 x 400 px² (Pixel size: 84 nm)
YOLOv2 model
The YOLOv2 model was generated using the ZeroCostDL4Mic platform (Chamier et al., 2021). It was trained from scratch for 97 epochs on 153 manually annotated images (image dimensions: (400, 400, 3)) with a batch size of 16 and a custom loss function combining MSE and crossentropy losses, using the YOLOv2 ZeroCostDL4Mic notebook (v 1.12) (von Chamier & Laine et al., 2020). Key python packages used include tensorflow (v0.1.12), Keras (v 2.3.1), numpy (v 1.19.5), cuda (v 10.1.243). The training was accelerated using a Tesla P100GPU and data was augmented by a factor of 8 using rotation and flipping.
The model weights can be used with the ZeroCostDL4Mic YOLOv2 notebook.
Author(s): Christoph Spahn1,2, Mike Heilemann1,3
Contact email: christoph.spahn@mpi-marburg.mpg.de
Affiliation(s):
1) Institute of Physical and Theoretical Chemistry, Max-von-Laue Str. 7, Goethe-University Frankfurt, 60439 Frankfurt, Germany
2) ORCID: 0000-0001-9886-2263
3) ORCID: 0000-0002-9821-3578