New paper: Deep Permutation Equivariant Structure from Motion. We propose an architecture that, given a set of point tracks in multiple images of a scene, recovers both the camera params and 3D structure by minimizing an unsupervised reprojection loss arxiv.org/abs/2104.06703

You need to read 8 #ComputerVision papers a day before a top tier conference serves you the next batch. Care about #MachineLearning, too? Read 23 a day. Keep up!

The code for our paper: "Layered Neural Atlases for Consistent Video Editing" is now available! github.com/ykasten/layere…

Check out the code for our ICCV 2021 oral presentation paper "Deep Permutation Equivariant Structure from Motion"! Hodaya Koslowsky Yoni Kasten Haggai Maron Meirav Galun Ronen Basri github.com/drormoran/Equi…

The Feinberg Graduate School invites #undergraduate, #graduate students, and #postdoc researchers to apply to our supported Visiting Students Program for #Ukrainians: 5K NIS per month up to 6 months + accommodation + airfare. For more info 👉 bit.ly/3HNofdw #ukraine

Our exciting new paper applies spectral analysis to the neural tangent kernel to unravel the implicit bias of deep, over-parametrized *CNNs*, revealing advantages over fully-connected networks and underscoring the role of hierarchy. arxiv.org/abs/2203.09255

Agreed! Amazing feat of engineering. Would be fun to teach a course on robotics with this.

The workshop below will feature my talk about recent advances in *GNN expressivity and subgraph networks* as well as many other highly interesting talks. Do not miss it! Organized by Efi Tsamoura and Martin Grohe and Dr Vaishak Belle

Checkout our new SIGGRPAH paper! A simple self-distillation approach is all you need to make StyleGAN capable of generating high quality images from noisy Internet collections

Congratulations to my PhD student Peter Hubry and his co-authors Timothy Duff, Anton Leykin and Tomas Pajdla for winning the CVPR 2022 Best Paper Award! #cvpr2022 lnkd.in/eQAwbzTc

EU: driving assistance systems are now mandatory for all vehicles (all based on deep learning). Also EU: All safety-critical AI systems must be explainable (deep learning must be banned).

In our new ICML paper, we introduce CNFM: Continuous Normalizing Flow Matching. A new approach for scalable CNFs on manifolds, the first generative model to scale to moderately high dimensional manifolds! arxiv.org/abs/2207.04711 (1/n)

In 1945 we defeated violent fascist leaders who controlled the press & public perceptions of truth. If you were 20 then, you are now 97 — or dead. I wonder if dictators are popular today because hardly anyone alive remembers the havoc they wrought & why we fought against them.

Our paper on subgraph GNNs was selected for an oral presentation at #NeurIPS22! with Beatrice Bevilacqua Fabrizio Frasca Michael Bronstein

Go, Liverpool! theguardian.com/science/2022/o…

The AI explosion is warping our sense of time. Can you believe Stable Diffusion is only 4 months old, and ChatGPT <4 weeks old 🤯? If you blink, you miss a whole new industry. Here are my TOP 10 AI spotlights, from a breathtaking 2022 in rewind ⏮: a long thread 🧵

From the GPT-4 paper. Not sure how I feel about this.

Self-supervised representation learning (DINO) + test time optimization=DINO-tracker! Achieving SOTA tracking results across long range occlusions. Congrats Narek Tumanyan Assaf Singer and OneViTaDay on the great work! 🦖🦖

מלחמת התנומה