Sean
@sean_8100
Ph.D. student @TechnionLive under Prof. Michael Elad ; Researching Image Inverse problems ; sean-man 🦋
optimization theorem: "assume a lipschitz constant L..." the lipschitz constant:
[1/9] We created a performant Lipschitz transformer by spectrally regulating the weights—without using activation stability tricks: no layer norm, QK norm, or logit softcapping. We think this may address a “root cause” of unstable training.
I randomly happened upon a recording of a talk I gave at Stanford in late 2019. It's a good snapshot of some of the computational photography work I was doing from 2013-2019 (denoising, white balance, tone mapping, portrait mode). Fun research decade in retrospect!
📷 FlowEdit has been accepted to @ICCVConference Edit real images with text-to-image flow models! Check out: code github.com/fallenshock/Fl… webpage matankleiner.github.io/flowedit/ space to edit your images - huggingface.co/spaces/fallens… great ComfyUI plugins (@logtdx) matankleiner.github.io/flowedit/#comfy
If you're at @icmlconf, come meet me at the posters this morning (11-13:30) presenting "When Diffusion Models Memorize" @ Hall A-B #E-2109, and DDCM - "Compressed Image Generation with Denoising Diffusion Codebook Models" in the afternoon (16:30-19) @ Hall A-B #E-3105!
Excited to present DDCM📖 with @guy__ohayon, @t_michaeli & M. Elad🎉 DDCM is a new generative approach that achieves SotA compression and is extendable to compressed conditional synthesis, all just w/ pre-trained diffusion models. Webpage &🤗Demo: ddcm-2025.github.io
Happy to announce our latest paper: SingLoRA! We replace the usual LoRA formulation (BA) with a single matrix (AᵀA), leading to better optimization, stability, and improved performance across tasks. Check it out! Thanks for sharing, @_akhaliq!
SingLoRA Low Rank Adaptation Using a Single Matrix
Excited to announce my work during @GoogleAI internship is now on arXiv! Grateful for my incredible hosts and collaborators: @2ptmvd, @docmilanfar, @KangfuM, Mojtaba Sahraee-Ardakan and @ukmlv. Please check out the paper here: [arxiv.org/abs/2507.05604].
For better consistency and fewer artifacts in im2im diffusion models, consider Kernel Density Steering (KDS): It is inference time and training-free; uses an N-particle ensemble to derive an empirical density, & explicit mode-seeking to push samples to high-density regions 1/2
🚨Happy to share our new work "Guidance in the Frequency Domain Enables High-Fidelity Sampling at Low CFG Scales" TLDR: Applying classifier-free guidance in the frequency domain boosts quality at low CFG scales, while avoiding the issues of high CFG scales by design. 🧵👇
Iranians, check on your Israeli friends and colleagues. Israelis, check on your Iranian friends and colleagues. Because what will bring lasting peace starts and ends with the personal connections, the love and affection many of us already have for each other.
We attempted to make Normalizing Flows work really well, and we are happy to report our findings in paper arxiv.org/pdf/2412.06329, and code github.com/apple/ml-tarfl…. [1/n]
🧵 What if a single UNet could solve all inverse problems? In our latest preprint with @HuraultSamuel, Maxime Song, and @TachellaJulian, we build a single multitask UNet for computational imaging — and show it generalizes surprisingly well 👇 arxiv.org/abs/2503.08915
we've been working on this for a while, and i find it very useful. try it out. (consider this to be v0, more to come!)
Meet Ai2 Paper Finder, an LLM-powered literature search system. Searching for relevant work is a multi-step process that requires iteration. Paper Finder mimics this workflow — and helps researchers find more papers than ever 🔍
🚀 Do you want to learn about self-supervised learning for inverse problems? ▶️ Check out the 3-hour tutorial presented by Mike Davies and myself, covering recent advances in a unified and simplified framework! youtube.com/playlist?list=…
Excited to present DDCM📖 with @guy__ohayon, @t_michaeli & M. Elad🎉 DDCM is a new generative approach that achieves SotA compression and is extendable to compressed conditional synthesis, all just w/ pre-trained diffusion models. Webpage &🤗Demo: ddcm-2025.github.io
New paper: Generative Photography generative-photography.github.io/project/ We overcome a limit in today's generative tools by enabling camera controls. We are one step closer to generative models for professional photography. To appear @CVPR
Our paper "Function-Space Learning Rates" is on arXiv! We give an efficient way to estimate the magnitude of changes to NN outputs caused by a particular weight update. We analyse optimiser dynamics in function space, and enable hyperparameter transfer with our scheme FLeRM! 🧵👇