Thrilled to announce that this week, I had the fantastic opportunity to present our paper 'Trade-offs in Fine-tuned Diffusion Models Between Accuracy and Interpretability' as an oral presentation at @RealAAAI as part of the main conference - Safe, Robust, and Responsible AI Track
🚨Our latest work on data generation, privacy preservation and surrogate datasets has been (early-)accepted at #MICCAI24 !
And the generation process 120x faster than last year ! 🚀
🌟EchoNet-Synthetic: Privacy-preserving Video Generation for Safe Medical Data Sharing 🌟
🚀🩻 Exciting collaboration alert! The IDEA Lab @aibe_fau@UniFAU and pediatric specialists at University Hospital Erlangen (#UKER) joined forces to develop a new AI tool for assessing Fontan circulation complications in children with CHD. 🏥🖥️
@BorderlessSci@BernhardKainz1
Thrilled to announce that this week, I had the fantastic opportunity to present our paper 'Trade-offs in Fine-tuned Diffusion Models Between Accuracy and Interpretability' as an oral presentation at @RealAAAI as part of the main conference - Safe, Robust, and Responsible AI Track
To train interpretable text-to-image diffusion models, we find three main design principles that you should follow:
1. Keep the language encoder frozen
2. Keep the variability of your input text high
3. Minimize the domain gap
The 4th International Workshop of Advances in Simplifying Medical UltraSound (ASMUS) - a workshop held in conjunction with #MICCAI23 is only a few days away. Join us on
Sunday 8⃣ Oct
from 8⃣ AM in
Meeting Room 8⃣
Vancouver Convention Center East Building Level 1.
🔎 We've developed an improved self-supervised single-class training strategy that approximates probabilistic inference with loosen feature locality constraints. Our approach with Probabilistic Poisson Image Interpolation (P-PII) shows promise in identifying unknown unknowns.
Excited to go to #ICCV2023 next week, where I will present our work on "Foreground-Background Separation through Concept Distillation from Generative Image Foundation Models" during the poster session on Wednesday at 10:30 a.m. https://t.co/CeWo2kmeyj
If you are interested in how we leverage pre-trained diffusion models to teach them concepts like foreground and background, come visit me for some exciting discussions.
I am looking for a PhD student to be hosted in Germany @BorderlessSci, co-funded and co-located with industry. PM/email in case of interest (please attach your CV)
📢fully-funded PhD/PostDoc position in my group. Many opportunities for making progress in self-supervised learning for medical image analysis and beyond. Details of positions and how to apply here: https://t.co/jHQml19URC