Judy Hoffman

Stanford, Computer Vision Postdoctoral Researcher


New! Clockwork Convnets for Video Semantic Segmentation
Evan Shelhamer*, Kate Rakelly*, Judy Hoffman*, Trevor Darrell
Video Semantic Segmentation Workshop at European Conference in Computer Vision (ECCV), 2016.
*Authors Contributed Equally.

We reduce latency and increase throughput for video semantic segmentation by exploiting the stationarity of the upper level semantic activations of FCNs.

Learning with Side Information through Modality Hallucination
Judy Hoffman, Saurabh Gupta, Trevor Darrell
In Proc. Computer Vision and Pattern Recognition (CVPR), 2016.

A method to hallucinate mid-level activations for a missing modality at test time.

Cross Modal Distillation for Supervision Transfer
Saurabh Gupta, Judy Hoffman, Jitendra Malik
In Proc. Computer Vision and Pattern Recognition (CVPR), 2016.
bibtex / models and code

We propose a method for pre-training a deep network for a new imaging modality which lacks sufficient supervised training data.

Cross-Modal Adaptation for RGB-D Detection
Judy Hoffman, Saurabh Gupta, Jian Leong, Sergio Guadarrama, Trevor Darrell,
IEEE International Conference on Robotics and Automation (ICRA), 2016.

We propose a technique to adapt CNN based object detectors trained on RGB images to effectively leverage depth images at test time to boost detection performance.

Quantification in-the-wild: data-sets and baselines
Oscar Beijbom, Judy Hoffman, Evan Yao, Trevor Darrell, Alberto Rodriguez-Ramirez, Manuel Gonzlez-Rivero, Ove Hoegh-Guldberg.
Transfer and Multi-Task Learning: Trends and New Perspectives, Workshop at NIPS, 2015.

Introduces two new ecological datasets for domain adaptation for quantification.

Simultaneous Deep Transfer Across Domains and Tasks
Eric Tzeng*, Judy Hoffman*, Trevor Darrell, Kate Saenko
International Conference on Computer Vision (ICCV), 2015.
*Equal Contribution
bibtex / caffe branch / prototxt

We introduce a domain confusion and softlabel loss to simultaneously learn a visual representation which is both discriminative and renders the domains indistinguishable.

Spatial Semantic Regularisation for Large Scale Object Detection
Damian Mrowca, Marcus Rohrbach, Judy Hoffman, Ronghang Hu, Kate Saenko, Trevor Darrell
International Conference on Computer Vision (ICCV), 2015.

We propose a multi-class spatial regularization method based on adaptive affinity propagation clustering which simultaneously optimizes across all categories and all proposed locations in the image to improve both location and categorization of selected detection proposals.

Detector Discovery in the Wild: Joint Multiple Instance and Representation Learning
Judy Hoffman, Deepak Pathak, Trevor Darrell, Kate Saenko
Computer Vision and Pattern Recognition (CVPR), 2015.

We propose a model that simultaneously trains a representation and detectors for categories with either image-level or bounding-box localized labels present. We provide a novel formulation of a joint multiple instance learning method that combines the heterogenous data sources.

LSDA: Large Scale Detection through Adaptation
Judy Hoffman, Sergio Guadarrama, Eric Tzeng, Ronghang Hu, Jeff Donahue, Ross Girshick, Trevor Darrell, Kate Saenko
Neural Information Processing Symposium (NIPS), 2014.
bibtex / project page

Released >7.5K detector! We present a method to transform classifiers into detectors by transferring knowledge from known detector categories.

Continuous Manifold Based Adaptation for Evolving Visual Domains
Judy Hoffman, Trevor Darrell, Kate Saenko
Computer Vision and Pattern Recognition (CVPR), 2014.
bibtex / video / project page

We propose a method for adapting to unlabeled data over time by modeling a continuosly evolving domain.

Interactive Adaptation of Real-Time Object Detectors
Daniel Goehring, Judy Hoffman, Erik Rodner, Kate Saenko, Trevor Darrell
International Conference in Robotics and Automation (ICRA), 2014.
bibtex / project page

We propose a method for quickly training detectors for novel categories on in-situ image data.

DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition
Jeff Donahue, Yangqing Jia, Oriol Vinyals, Judy Hoffman, Ning Zhang, Eric Tzeng, Trevor Darrell
International Conference in Machine Learning (ICML), 2014.
bibtex / code

We propose a new feature based on deep convolutional neural networks and show improvement over state-of-the-art visual feature representations.

Asymmetric and Category Invariant Feature Transformations for Domain Adaptation
Judy Hoffman, Erik Rodner, Jeff Donahue, Brian Kulis, Kate Saenko
International Journal of Computer Vision, Special Domain Adaptation Addition, 2013.

Efficient Learning of Domain-invariant Image Representations
Judy Hoffman, Erik Rodner, Jeff Donahue, Kate Saenko, Trevor Darrell
International Conference on Learning Representations (ICLR), 2013. (Oral)
bibtex / talk / code

We learn a category invariant feature transformation, which maps target points into the source domain such that they corrected classified by the source classifier.

Semi-Supervised Domain Adaptation with Instance Constraints
Jeff Donahue, Judy Hoffman, Erik Rodner, Kate Saenko, Trevor Darrell
Computer Vision and Pattern Recognition (CVPR), 2013.
bibtex / poster

By using instance constraints, available through tracking or other methods, we can improve unsupervised domain adaptation performance.

Discovering Latent Domains For Multisource Domain Adaptation
Judy Hoffman, Brian Kulis, Trevor Darrell, Kate Saenko
European Conference in Computer Vision (ECCV), 2012.
supplementary material / bibtex / poster / video / code

We learn to separate large heterogeneous data sources into multiple latent visual domains and show that using this learned clustering improves classification performance.

Weakly Supervised Learning of Object Segmentations from Web-Scale Video
Glen Hartmann, Matthias Grundmann, Judy Hoffman, David Tsai, Vivek Kwatra, Omid Madani, Sudheendra Vijayanarasimhan, Irfan Essa, James Rehg, Rahul Sukthankar
European Conference in Computer Vision (ECCV) Workshop on Web-scale Vision and Social Media, 2012. (Best Paper Award)

We learn segment level video classification using videos with only weakly labeled tag information.

Domain Adaptation with Multiple Latent Domains
Judy Hoffman, Kate Saenko, Brian Kulis, Trevor Darrell
NIPS Domain Adaptation Workshop Talk, 2011. (Best Student Paper Award)

We present a method for multi-source adaptation with latent source domains. See ECCV2012 paper for more details.

EG-RRT: Environment-Guided Random Trees for Kinodynamic Motion Planning with Uncertainty and Obstacles
Leonard Jaillet, Judy Hoffman, Jur van den Berg, Pieter Abbeel, Josep M. Porta, Ken Goldberg
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2011.