T
Trevor Darrell
Researcher at University of California, Berkeley
Publications - 734
Citations - 222973
Trevor Darrell is an academic researcher from University of California, Berkeley. The author has contributed to research in topics: Computer science & Object detection. The author has an hindex of 148, co-authored 678 publications receiving 181113 citations. Previous affiliations of Trevor Darrell include Massachusetts Institute of Technology & Boston University.
Papers
More filters
Proceedings ArticleDOI
Modular Networks for Compositional Instruction Following.
TL;DR: This work proposes a modular architecture for following natural language instructions that describe sequences of diverse subgoals and finds that modularization improves generalization to novel subgoal compositions, as well as to environments unseen in training.
Proceedings ArticleDOI
Exposing the Limits of Video-Text Models through Contrast Sets
TL;DR: An evaluation framework is proposed that probes video-text models with hard negatives with high accuracy, and automatically builds contrast sets, where true textual descriptions are manipulated in ways that change their semantics while maintaining plausibility.
Journal ArticleDOI
Multitask Vision-Language Prompt Tuning
Sheng Shen,Shi-Yong Yang,Tianjun Zhang,Bohan Zhai,Joseph E. Gonzalez,Kurt Keutzer,Trevor Darrell +6 more
TL;DR: In this paper , the authors propose multitask vision-language prompt tuning (MVLPT), which incorporates cross-task knowledge into prompt tuning for vision-languages models, and demonstrate the effectiveness of learning a single transferable prompt from multiple source tasks to initialize the prompt for each target task.
Posted Content
Revisiting Few-shot Activity Detection with Class Similarity Control.
TL;DR: This paper presents a conceptually simple and general yet novel framework for few-shot temporal activity detection based on proposal regression which detects the start and end time of the activities in untrimmed videos.
Posted Content
Prototypical Cross-domain Self-supervised Learning for Few-shot Unsupervised Domain Adaptation
Xiangyu Yue,Zangwei Zheng,Shanghang Zhang,Yang Gao,Trevor Darrell,Kurt Keutzer,Alberto Sangiovanni Vincentelli +6 more
TL;DR: In this article, an end-to-end Prototypical Cross-domain Self-Supervised Learning (PCS) framework is proposed for few-shot domain adaptation, which not only performs cross-domain low-level feature alignment but also encodes and aligns semantic structures in the shared embedding space across domains.