Bioimage Analysis: Practice Deep Learning Without Coding
A. Badoual, D. Sage
Proceedings of the Ninth Edition of the CNRS Thematic School, Functional Microscopy for Biology (MiFoBio'21), Presqu'île de Giens, French Republic, November 5-12, 2021, pp. 138.
The emergence of deep learning techniques (a.k.a. neural networks) has drastically transformed the field of bioimage analysis and its quest for the understanding of biological processes. Successful use of these learning-based techniques in various biological imaging problems are found on a regular basis and the trend is likely to continue. Unfortunately, the deployment of deep learning models is often riddled with technical challenges for non-expert users, and their appropriate use requires deep learning knowledge and good programming skills. Since 2020, efforts have been made to democratize the use of deep learning with the deployment of notebooks, zoos of pre-trained models and plugins such as DeepImageJ. However, these user-friendly and code-free tools deserve to be better known and disseminated in the biology community.
The goal of this workshop is to contribute to the spread and assessment of deep learning models in life-sciences applications and bioimage informatics. First, we will get an intuitive understanding of deep learning concepts. Then, we will use pre-trained models with DeepImageJ, and finally, we will train neural networks without progamming for image segmentation. For this purpose, we will exploit with DeeplmageJ the pre-trained models gathered on the Bioimage Model Zoo. We will also use the notebooks developed by ZeroCostDL4Mic to train neural networks on google Colab. We rely on our long experience in teaching image processing on ImageJ for Master students to guide the biologists throughout the workshop. At the end of this workshop the participants will know how to choose the appropriate pre-trained model according to their application and how to test it on their own images. They will also leave the workshop with basic knowledge on how to train a model from scratch if needed.
This workshop is designed for participants without any skills in programmation and without machine-learning knowledge.
@INPROCEEDINGS(http://bigwww.epfl.ch/publications/badoual2102.html,
AUTHOR="Badoual, A. and Sage, D.",
TITLE="Bioimage Analysis: {P}ractice Deep Learning Without Coding",
BOOKTITLE="Proceedings of the Ninth Edition of the {CNRS} Thematic
School, Functional Microscopy for Biology ({MiFoBio'21})",
YEAR="2021",
editor="",
volume="",
series="",
pages="138",
address="Presqu'{\^{i}}le de Giens, French Republic",
month="November 7-12,",
organization="",
publisher="",
note="")