Many papers have truly interesting ideas, but instead of presenting them clearly, they mix them with all kinds of other params and models to get SOTA, like: ours = okay ours + baseline + optimizer X = sota!!! Can we just agree that a paper can be interesting w/o SOTA?

The curated list of Q&A's for our #StarDist webinar @NEUBIAS_COST can now be found at the forum: forum.image.sc/t/neubias-acad… Webinar: youtube.com/watch?v=Amn_eH… Big thanks to Olivier Burri, Dr Siân Culley and esp. Uwe Schmidt!

New Python releases for #CSBDeep and #StarDist! Lots of improvements under the hood, including support for TensorFlow 2. Joint effort with Martin Weigert (maweigert.bsky.social)! All changes: github.com/CSBDeep/CSBDee… github.com/mpicbg-csbd/st…

New back-to-back releases of our DL python packages csbdeep and stardist. Among many other things, now allows for using tf2, stitched predictions on large 2D images, and importing 3D stardist results in Blender :) kudos to Uwe Schmidt!

Good benchmarks are important for progress in ML, but personally I’m more interested in research that explore new methods capable of doing things that existing SOTA methods simply fail at doing. I also think pursuing such research is easier and more enjoyable than beating SOTA.

Reviewers automatically assume that simple is not novel. This is sheer laziness. Yes, it may be simple and obvious in retrospect, but someone had to have that insight first. Simple is good. Simple is robust, easy to implement and reproduce, broadly applicable, etc.

Now live @nature - on a new mechanism of nuclear remodelling in closed mitosis! If you missed the bioRxiv or want to know what changed, thread to follow - once my co-authors & I finish our socially distanced pint 🍺 nature.com/articles/s4158…

I'm a Research Software Engineer looking for challenging job opportunities since the lab is relocating in a couple of months. Dresden or remote. Experienced in open source development / image analysis / data visualization / deep learning / human computer interaction / ..

Interested in using neural networks for estimating optical aberrations without collecting labelled ground truth data? Try out our approach - PhaseNet. Martin Weigert (maweigert.bsky.social) Uwe Schmidt Eugene W Myers Jr paper - osapublishing.org/oe/abstract.cf… code - github.com/mpicbg-csbd/ph…

#I2K. 2x4 hour tutorials on #ImgLib2 #BDV #Numpy #Weka #CellPose #N5 #ImageJ2 #Ilastik #CLIJ #ImJoy #mManager #OMERO #ITK #SciView #CellProfiler #Icy #CSBDeep #CARE #StarDist #Noise2Void #BigStitcher Missing something? Respond to invite or apply. Oct 25th janelia.org/you-janelia/co…

Behold SplineDist! biorxiv.org/content/10.110…

Our FIB-SEM segmentation paper revealing part of the inner workings of insulin producing beta cells is out Journal of Cell Biology! Great and super enjoyable teamwork under the lead of @Andi_Microscopy with Frau Zufall @msolimena HHMI | Janelia @PLID_info MPI-CBG Dresden Florian Jug! EPFL Center for Imaging

We just made a new stardist release, one of the bigger ones! Multi-class prediction, memory/runtime gains, and integration for the shiny new napari plugin :) As usual, great colab with Uwe Schmidt and many others: github.com/stardist/stard…

Our new paper on "The Bayesian Learning Rule" is now on arXiv, where we provide a common learning-principle behind a variety of learning algorithms (optimization, deep learning, and graphical models). arxiv.org/abs/2107.04562 Guess what, the principle is Bayesian. A very long🧵

I get a lot of reviews that say my work is not novel and I bet I'm not alone. It's always frustrating because I see novelty where the reviewer doesn't. Rather than rebut every critique, I've written a blog post to help reviewers think about novelty. perceiving-systems.blog/en/news/novelt…

How well does #StarDist work on histopathology images? Pretty well! Our StarDist approach (with Uwe Schmidt) just won the nuclei seg/class task of the ISBI 2022 CoNIC challenge! 🎉 Paper: arxiv.org/abs/2203.02284 conic-challenge.grand-challenge.org/evaluation/seg…

Together with the wonderful collaborators and community partners, we are happy to finally present the BioImage Model Zoo at bioimage.io, now described in biorxiv.org/content/10.110….

If you're interested in the details of our #StarDist-based submission to the 2022 CoNIC challenge, we finally released the code. github.com/stardist/stard…