Derek Hoiem

papers

1,458

total citations

papers (13)

Dreaming to Distill: Data-Free Knowledge Transfer via DeepInversion

CVPR 2020arXiv

665

citations

Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision Language Audio and Action

CVPR 2024arXiv

280

citations

Language Models with Image Descriptors are Strong Few-Shot Video-Language Learners

NEURIPS 2022arXiv

164

citations

RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations

CVPR 2025arXiv

citations

REN: Fast and Efficient Region Encodings from Patch-Based Image Encoders

NEURIPS 2025arXiv

citations

Towards General Purpose Vision Systems: An End-to-End Task-Agnostic Vision-Language Architecture

CVPR 2022

citations

Derek Hoiem

papers (13)

Dreaming to Distill: Data-Free Knowledge Transfer via DeepInversion

Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision Language Audio and Action

Language Models with Image Descriptors are Strong Few-Shot Video-Language Learners

Contrastive Learning for Weakly Supervised Phrase Grounding

Webly Supervised Concept Expansion for General Purpose Vision Models

StyleGAN knows Normal, Depth, Albedo, and More

PatchMatch-RL: Deep MVS With Pixelwise Depth, Normal, and Visibility

Region-Based Representations Revisited

Improving Confidence Estimates for Unfamiliar Examples

Anytime Continual Learning for Open Vocabulary Classification

RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations

REN: Fast and Efficient Region Encodings from Patch-Based Image Encoders

Towards General Purpose Vision Systems: An End-to-End Task-Agnostic Vision-Language Architecture

papers (13)

Dreaming to Distill: Data-Free Knowledge Transfer via DeepInversion

Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision Language Audio and Action

Language Models with Image Descriptors are Strong Few-Shot Video-Language Learners

Contrastive Learning for Weakly Supervised Phrase Grounding

Webly Supervised Concept Expansion for General Purpose Vision Models

StyleGAN knows Normal, Depth, Albedo, and More

PatchMatch-RL: Deep MVS With Pixelwise Depth, Normal, and Visibility

Region-Based Representations Revisited

Improving Confidence Estimates for Unfamiliar Examples

Anytime Continual Learning for Open Vocabulary Classification

RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations

REN: Fast and Efficient Region Encodings from Patch-Based Image Encoders

Towards General Purpose Vision Systems: An End-to-End Task-Agnostic Vision-Language Architecture