TY - JOUR
T1 - Democratising deep learning for microscopy with ZeroCostDL4Mic
AU - von Chamier, Lucas
AU - Laine, Romain F
AU - Jukkala, Johanna
AU - Spahn, Christoph
AU - Krentzel, Daniel
AU - Nehme, Elias
AU - Lerche, Martina
AU - Hernández-Pérez, Sara
AU - Mattila, Pieta K
AU - Karinou, Eleni
AU - Holden, Séamus
AU - Solak, Ahmet Can
AU - Krull, Alexander
AU - Buchholz, Tim-Oliver
AU - Jones, Martin L
AU - Royer, Loïc A
AU - Leterrier, Christophe
AU - Shechtman, Yoav
AU - Jug, Florian
AU - Heilemann, Mike
AU - Jacquemet, Guillaume
AU - Henriques, Ricardo
PY - 2021/12
Y1 - 2021/12
N2 - Deep Learning (DL) methods are powerful analytical tools for microscopy and can outperform conventional image processing pipelines. Despite the enthusiasm and innovations fuelled by DL technology, the need to access powerful and compatible resources to train DL networks leads to an accessibility barrier that novice users often find difficult to overcome. Here, we present ZeroCostDL4Mic, an entry-level platform simplifying DL access by leveraging the free, cloud-based computational resources of Google Colab. ZeroCostDL4Mic allows researchers with no coding expertise to train and apply key DL networks to perform tasks including segmentation (using U-Net and StarDist), object detection (using YOLOv2), denoising (using CARE and Noise2Void), super-resolution microscopy (using Deep-STORM), and image-to-image translation (using Label-free prediction - fnet, pix2pix and CycleGAN). Importantly, we provide suitable quantitative tools for each network to evaluate model performance, allowing model optimisation. We demonstrate the application of the platform to study multiple biological processes.
AB - Deep Learning (DL) methods are powerful analytical tools for microscopy and can outperform conventional image processing pipelines. Despite the enthusiasm and innovations fuelled by DL technology, the need to access powerful and compatible resources to train DL networks leads to an accessibility barrier that novice users often find difficult to overcome. Here, we present ZeroCostDL4Mic, an entry-level platform simplifying DL access by leveraging the free, cloud-based computational resources of Google Colab. ZeroCostDL4Mic allows researchers with no coding expertise to train and apply key DL networks to perform tasks including segmentation (using U-Net and StarDist), object detection (using YOLOv2), denoising (using CARE and Noise2Void), super-resolution microscopy (using Deep-STORM), and image-to-image translation (using Label-free prediction - fnet, pix2pix and CycleGAN). Importantly, we provide suitable quantitative tools for each network to evaluate model performance, allowing model optimisation. We demonstrate the application of the platform to study multiple biological processes.
KW - Animals
KW - Cell Line, Tumor
KW - Cloud Computing
KW - Datasets as Topic
KW - Deep Learning
KW - Humans
KW - Image Processing, Computer-Assisted/methods
KW - Microscopy/methods
KW - Primary Cell Culture
KW - Rats
KW - Software
U2 - 10.1038/s41467-021-22518-0
DO - 10.1038/s41467-021-22518-0
M3 - Article
C2 - 33859193
SN - 2041-1723
VL - 12
JO - Nature Communications
JF - Nature Communications
IS - 1
M1 - 2276
ER -