Raphael Schäfer
@Telcrome
Followers
30
Following
186
Media
0
Statuses
16
Research Engineer @FraunhoferMEVIS in Aachen
Joined June 2012
RT @FraunhoferMEVIS: #PressRelease: MEVIS develops a multitasking approach to train foundation models efficiently with minimal data. Data s….
0
2
0
RT @FraunhoferMEVIS: New paper “Overcoming data scarcity in biomedical imaging with a foundational multi-task model in @NatComputSci: https….
0
3
0
RT @FraunhoferMEVIS: New #preprint “Overcoming Data Scarcity in Biomedical Imaging with a Foundational Multi-Task Model”: .
0
3
0
🔜 Code coming soon! Thanks to the team, my colleagues and the countless people who shared labeled data and made large-scale supervised training possible! 🩺💻👩💻 #OpenScience #FoundationalModels.
0
0
0
🚀 UMedPT outperforms traditional ImageNet pretraining and provides a new base model for future research. It maintains performance with just 1% of the data for in-domain tasks, and 50% for out-of-domain tasks. This is a game changer for research with limited datasets. (4/n).
1
0
1
Here it was used it to train UMedPT. UMedPT was jointly trained for classification, segmentation and object detection. Rather than requiring the collection of a large, exhaustively labeled dataset, UMedPT incorporated public tomographic, microscopic and X-ray data as is! (3/n).
1
0
0
Our research introduces a new multitask training strategy for medical pretraining, which can train on a large number of tasks while leveraging the original labels of each task. (2/n).
1
0
0
🤖🔬 I am excited to announce our latest work on foundational models for medical imaging @FraunhoferMEVIS !.(1/n) 🧵. @n_till41526, Henning Höfener, Annkristin Lange, Dorit Merhof, Friedrich Feuerhake, @SchulzVolkmar @johannes_lotz @Kiessling_F.
1
2
3
RT @johannes_lotz: Topic of my talk today: Robustness of AI in pathology can be achieved by pre-training medical „foundational“ models with….
0
4
0
RT @Karl_Lauterbach: In Deutschland werden leider oft die Daten besser geschützt als die Patienten. Ohne Forschungsdaten können wir uns zB….
www.tagesschau.de
Mediziner, Politiker und Patienten wollen Gesundheitsdaten stärker für Forschungszwecke nutzen. Doch bei der Umsetzung hapert es - woran liegt das? Auf einem Kongress beraten Experten über Lösungen....
0
46
0
RT @johannes_lotz: Our Tissue Concepts model scored second at the #SemiCol challenge on detection and segmentation of colorectal cancer! 🥳….
0
2
0
RT @FraunhoferMEVIS: Improved diagnostic performance by multitask pretraining from heterogeneous medical data—find out more in our e-poster….
0
2
0
RT @FraunhoferMEVIS: Mit dezentraler #KI die Behandlung von #Prostatakrebs verbessern: #Forschungsprojekt PROSurvival gestartet: https://t.….
0
4
0
RT @johannes_lotz: I am happy to talk about image registration in pathology and how it is useful to train AI at the seminar series @TIAwarw….
0
1
0
RT @FraunhoferMEVIS: How can clinicians and patients choose the most suitable treatment? Read more about what our researchers are working o….
0
3
0