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
Journal ArticleDOI
Refocusing Is Key to Transfer Learning
TL;DR: To p-down a tention st eering (TOAST) as mentioned in this paper is a novel transfer learning algorithm that keeps the pre-trained backbone frozen, while selecting the taskrelevant elements in the output and feeding them back to the model to steer its attention to the task-specific features.
Posted Content
Confidence Adaptive Anytime Pixel-Level Recognition.
TL;DR: In this article, a cascade of "exits" is attached to the model to make multiple predictions and direct further computation, and a novel spatially adaptive approach is developed to avoid further computation on regions where early predictions are already sufficiently confident.
Posted Content
Task-Aware Deep Sampling for Feature Generation.
TL;DR: This work proposes a sample efficient learning model composed of a TDS generator, a discriminator and a classifier (e.g., a soft-max classifier) and finds that this model achieves state-of-the-art results on the compositional zero- shot learning benchmarks as well as improving upon the established benchmarks in conventional zero-shot learning with a faster convergence rate.
Proceedings ArticleDOI
Teachable Reinforcement Learning via Advice Distillation
TL;DR: In puzzle-solving, navigation, and locomotion domains, it is shown that agents that learn from advice can acquire new skills with significantly less human supervision than standard reinforcement learning algorithms and often less than imitation learning.