Romain F. Laine Profile picture
MRC Research fellow, scientist, microscopist, biophysicist @MRC_LMCB / @TheCrick #SRRF'ing, #DeepLearning, studying #HIV and #flu. Views my own. He/Him

Apr 15, 2021, 12 tweets

We are pleased to announce that our #ZeroCostDL4Mic work is now published!
nature.com/articles/s4146…
And our @github
github.com/HenriquesLab/Z…
#DeepLearning and #Microscopy for all!
For the occasion, I’ve put a small Tweetorial together to explain what this is all about!
Thread 0/7

Thread 1/7

You can implement Deep Learning for microscopy in several ways:
1- Using local resources (fast GPUs)
2- Using cloud-computing
3- Using pre-trained models

We think that (1) is uncommon, and that (3) can be unreliable. So we do (2)!

Thread 2/7

We built a platform using @GoogleColab and @ProjectJupyter notebooks that can perform training, quality control and prediction, all on the cloud and for free!

All of this dedicated to #microscopy !

Thread 3/7

We implemented a large range of neural networks from some of the best Deep Learning developer teams in the world for segmentation, object detection, denoising, restauration, super-resolution, image-to-image translation, etc.

Thread 4/7

We provide important features such as Data augmentation and Transfer learning. These allow efficient training of all the networks we provide.

Thread 5/7

We put a strong emphasis on providing an easy way to assess model quality and performance. This is important to validate any model.

Thread 6/7

We could not do it on our own and we got help from so many great researchers! Part 1/2

@ChamierLucas
@LaineBioImaging
@Liehuletti
@miCHRIScopy
@DanielKrentzel
@martina_lerche
@sara_mattilalab
@mattilalab
@Kar__El
@seamus_holden

Thread 7/7

We could not do it on our own and we got help from so many great researchers! Part 2/2

@_ahmetcansolak
@sagzehn
@tibuch_
@martinjones78
@loicaroyer
@christlet
@ShechtmanLab
@florianjug
Mike Heilemann
@guijacquemet
@HenriquesLab

Some highlights! #DeepSTORM from @ShechtmanLab with data from @christlet

#Noise2Void from @sagzehn with some nice live data from iSIM!

SI Thread 1/2
We also quantified the limitations of the approach (using @GoogleColab ) and present where it failed in our hands! Breaking points still allow us to train all our networks efficiently though.

SI Thread 2/2
We also showed that it's possible to use our notebooks on other cloud-based platforms such as @DeepnoteHQ and @FloydHub_ , others are available too !

Share this Scrolly Tale with your friends.

A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.

Keep scrolling