Amir Bar


I am a Ph.D. candidate at Tel Aviv University and UC Berkeley working with Amir Globerson and Trevor Darrell. My goal is to teach computers to understand the world from unlabeled data, using little to no supervision.

I've pioneered AI for medical imaging as an AI Research Lead at Zebra Medical Vision (acquired). My research team developed multiple FDA cleared algorithms for automatic analysis of medical images (e.g, [1, 2]).

Email  /  Twitter  /  GitHub  /  Google Scholar  /  LinkedIn  /  CV

  • Beginning August 2024, I will join Meta AI (FAIR) as a Postdoctoral Researcher, working with Yann LeCun.
  • I'm organizing the first workshop on Prompting in Vision at CVPR 24'. See you in Seattle!
  • I'm part of the organizing committee of NeurIPS 2023.

EgoPet: Egomotion and Interaction Data from an Animal's Perspective
Amir Bar, Arya Bakhtiar, Danny Tran, Antonio Loquercio, Jathushan Rajasegaran, Yann LeCun, Amir Globerson, Trevor Darrell
Technical Report, 2024
Project Page | Data | Code

We present EgoPet, a new egocentric video dataset of animals.

Finding Visual Task Vectors    
Alberto Hojel, Yutong Bai, Trevor Darrell, Amir Globerson, Amir Bar
Technical Report, 2024

We find "Task Vectors", latent activations that can guide models to perform computer vision tasks.

IMProv: Inpainting-based Multimodal Prompting for Computer Vision Tasks    
Jiarui Xu, Yossi Gandelsman, Amir Bar, Jianwei Yang, Jianfeng Gao, Trevor Darrell, Xiaolong Wang
Technical Report, 2023
Project Page | Code/Data | Demo

IMProv performs multimodal in-context learning to solve computer vision tasks.

Selected Publications

Stochastic positional embeddings improve masked image modeling    
Amir Bar, Florian Bordes, Assaf Shocher, Mahmoud Assran, Pascal Vincent, Nicolas Ballas, Trevor Darrell, Amir Globerson, Yann LeCun
ICML, 2024

Modeling location uncertainties via stochastic positional embeddings (StoP) improve masked image modeling.

Sequential Modeling Enables Scalable Learning for Large Vision Models
Yutong Bai, Xinyang Geng, Karttikeya Mangalam, Amir Bar, Alan Yuille, Trevor Darrell, Jitendra Malik, Alexei A Efros
CVPR, 2024
Project Page | Code

Large Vision Model trained on 420B tokens; exhibits interesting in-context learning capabilities.

Visual Prompting via Image Inpainting    
Amir Bar*, Yossi Gandelsman*, Trevor Darrell, Amir Globerson, Alexei A. Efros
NeurIPS, 2022
Project Page | Code/Data

Adapt a pre-trained visual model to novel downstream tasks without task-specific finetuning or any model modification.

* Equally contributed.
fast-texture Bringing Image Scene Structure to Video via Frame-Clip Consistency of Object Tokens
Elad Ben-Avraham, Roei Herzig, Karttikeya Mangalam, Amir Bar, Anna Rohrbach, Leonid Karlinsky, Trevor Darrell, Amir Globerson
NeurIPS, 2022
Project Page | Code
Incorporating image level scene structure during training improves video transformers.

Winner of the Ego4D CVPR'22 PNR Temporal Localization Challenge
fast-texture Object-Region Video Transformers
Roei Herzig, Elad Ben-Avraham, Karttikeya Mangalam, Amir Bar, Gal Chechik,
Anna Rohrbach, Trevor Darrell, Amir Globerson
CVPR, 2022
Project Page | Code

Incorporating objects into transformer layers improves video transformers.

DETReg: Unsupervised Pretraining with Region Priors for Object Detection
Amir Bar, Xin Wang, Vadim Kantorov, Colorado J Reed, Roei Herzig, Gal Chechik, Anna Rohrbach, Trevor Darrell, Amir Globerson
CVPR, 2022
Project Page | Code | Video | Demo

Pretraining transformers to localize potential objects improves object detection.

Compositional Video Synthesis with Action Graphs    
Amir Bar*, Roei Herzig*, Xiaolong Wang, Anna Rohrbach, Gal Chechik, Trevor Darrell, Amir Globerson
ICML, 2021
Project Page | Code | Video

We introduce Action Graphs, a structure that can better capture the compositional and hierrchical nature of actions. We propose a goal-oriented video synthesis task of *Action Graph to Video*

* Equally contributed.

Learning Canonical Representations for Scene Graph to Image Generation
Roei Herzig*, Amir Bar*, Huijuan Xu, Gal Chechik, Trevor Darrell, Amir Globerson
ECCV, 2020
Project Page | Code | Video

We present a model for Scene Graph to Image generation which is more robust to complex input scene graphs.

* Equally contributed.

Learning Individual Styles of Conversational Gesture  
Shiry Ginosar*, Amir Bar*, Gefen Kohavi, Caroline Chan, Andrew Owens, Jitendra Malik
CVPR, 2019
Press | Project Page | Code | Data

We predict plausible gestures to go along with someone's speech.

* Equally contributed.


The man allowed that about health captain played that alleged to Marks live up in the club comes the handed up moved to a brief

Language Generation with Recurrent Generative Adversarial Networks without Pre-training
Ofir Press*, Amir Bar*, Ben Bogin*, Jonathan Berant, Lior Wolf
1st Workshop on Learning to Generate Natural Language at ICML, 2017

We show that recurrent neural networks can be trained to generate text with GANs from scratch and vastly improve the quality of generated sequences compared to a convolutional baseline.

* Equally contributed.

3D Convolutional Sequence to Sequence Model for Vertebral Compression Fractures Identification in CT
David Chettrit, Tomer Meir, Hila Lebel, Mila Orlovsky, Ronen Gordon, Ayelet Akselrod-Ballin*, Amir Bar*
MICCAI, 2020
Press 1 2

We present a novel architecture used to detect vertebral compression fractures in Chest and Abdomen CT.

* Equally advised.

Automated opportunistic osteoporotic fracture risk assessment using computed tomography scans to aid in FRAX underutilization    
Noa Dagan, Eldad Elnekave, Noam Barda, Orna Bregman-Amitai, Amir Bar, Mila Orlovsky, Eitan Bachmat & Ran D. Balicer
Nature Medicine, 2020
Press 1 2

Methods for identifying patients at high risk for osteoporotic fractures are underutilized. We demonstrate it is feasibile to automatically evaluate risk based on routine abdomen or chest computed tomography (CT) scan.

Improved ICH classification using task-dependent learning  
Amir Bar, Michal Mauda-Havakuk, Yoni Turner, Michal Safadi, Eldad Elnekave
ISBI, 2019

Intracranial hemorrhage (ICH) is among the most critical and timesensitive findings to be detected on Head CT. We present a new architecture designed for optimal triaging of Head CTs, with the goal of decreasing the time from CT acquisition to accurate ICH detection. These results are comparable to previously reported results with smaller number of tagged studies.

Simulating Dual-Energy X-Ray Absorptiometry in CT Using Deep-Learning Segmentation Cascade
Arun Krishnaraj, Spencer Barrett, Orna Bregman-Amitai , Michael Cohen-Sfady, Amir Bar, David Chettrit, Mila Orlovsky, Eldad Elnekave
Journal of the American College of Radiology, 2019

Osteoporosis is an underdiagnosed condition despite effective screening modalities. The purpose of this study was to describe a method to simulate lumbar DEXA scores from routinely acquired CT studies using a machine-learning algorithm.

PHT-bot: a deep learning based system for automatic risk stratification of COPD patients based upon signs of pulmonary hypertension
David Chettrit, Orna Bregman Amitai, Itamar Tamir, Amir Bar and Eldad Elnekave
SPIE, 2019

Chronic Obstructive Pulmonary Disease (COPD) is a leading cause of morbidity and mortality worldwide. We apply deep learning algorithm to detect those at risk.

Compression Fractures Detection on CT
Amir Bar, Lior Wolf, Orna Bregman Amitai, Eyal Toledano, Eldad Elnekave
SPIE, 2017
Press 1 2

The presence of a vertebral compression fracture is highly indicative of osteoporosis and represents the single most robust predictor for development of a second osteoporotic fracture in the spine or elsewhere. We present an automated method for detecting spine compression fractures in Computed Tomography (CT) scans.

Systems and methods for automated detection of visual objects in medical images. US Patent 11,776,243
Cross modality training of machine learning models. US Patent 11,587,228.
Identification of a contrast phase depicted in a medical image. US Patent 11,727,087
Undergraduate and MA Researchers
Danny Tran
Arya Bakhtiar
Alberto Hojel

Template, Last update: 5/2024.