DeepBacs demonstrates the potential of open-source deep-learning approaches in microbiological research. We provide dataset that can be used to train models for different tasks, e.g. image segmentation, denoising, artificial labeling or prediction of super-resolution images.
In our opinion, the lack of meaningful examples and accessible image data prevents microbiologists to explore deep learning and thus experience its potential.
DeepBacs does not provide novel code, but uses popular DL networks implemented in the ZeroCostDL4Mic platform. It provides easy access to DL networks in a streamlined format and, together with suitable data, enables scientists to deep into the DL universe rapidly and without hardware acquistion.
List of tasks with links to the respective Wiki pages:
The learning approaches employed in DeepBacs are based on the work of many deep learning pioneers.
Further details about that can be found on the ZeroCostDL4Mic wiki and our DeepBacs wiki.
Christoph Spahn, Romain F. Laine, Pedro Matos Pereira, Estibaliz Gómez de Mariscal, Lucas von Chamier, Mia Conduit, Mariana Gomes de Pinho, Guillaume Jacquemet, Séamus Holden, Mike Heilemann, Ricardo Henriques. DeepBacs: Bacterial image analysis using open-source deep learning approaches. bioRxiv, 2021. DOI: https://doi.org/10.1101/2021.11.03.467152
Our work is based on the networks created by the amazing DL community. We are thus grateful and like to thank the people creating the networks used in DeepBacs:
- Martin Weigert
- Uwe Schmidt
- Florian Jug
- Alexander Krull
- Loic A. Royer
- Chawin Ounkomol
- Gregory R. Johnson
- Phillip Isola
- Alexei A. Efros
- Joseph Redmon
- Ali Farhadi