Yedid Hoshen

Affiliations

Hebrew University of Jerusalem

papers

517

total citations

papers (11)

PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation

CVPR 2021arXiv

299

citations

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion

ECCV 2024arXiv

citations

An Image is Worth More Than a Thousand Words: Towards Disentanglement in The Wild

NEURIPS 2021arXiv

citations

ObjectMate: A Recurrence Prior for Object Insertion and Subject-Driven Generation

ICCV 2025arXiv

citations

Deep Linear Probe Generators for Weight Space Learning

ICLR 2025arXiv

citations

Learning on Model Weights using Tree Experts

CVPR 2025arXiv

citations

Yedid Hoshen

Affiliations

papers (11)

PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion

An Image is Worth More Than a Thousand Words: Towards Disentanglement in The Wild

Membership Inference Attacks Are Easier on Difficult Problems

Scaling-Up Disentanglement for Image Translation

Image Shape Manipulation From a Single Augmented Training Sample

A Contrastive Objective for Learning Disentangled Representations

Recovering the Pre-Fine-Tuning Weights of Generative Models

ObjectMate: A Recurrence Prior for Object Insertion and Subject-Driven Generation

Deep Linear Probe Generators for Weight Space Learning

Learning on Model Weights using Tree Experts

papers (11)

PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion

An Image is Worth More Than a Thousand Words: Towards Disentanglement in The Wild

Membership Inference Attacks Are Easier on Difficult Problems

Scaling-Up Disentanglement for Image Translation

Image Shape Manipulation From a Single Augmented Training Sample

A Contrastive Objective for Learning Disentangled Representations

Recovering the Pre-Fine-Tuning Weights of Generative Models

ObjectMate: A Recurrence Prior for Object Insertion and Subject-Driven Generation

Deep Linear Probe Generators for Weight Space Learning

Learning on Model Weights using Tree Experts