Julieta Martinez, Michael J. Black, Javier Romero.
On human motion prediction using recurrent neural networks. In CVPR 2017 (29.84% acceptance rate)
We take a close look at deep recurrent approaches for human motion prediction, and propose a simple and scalable architecture that outperforms the state of the art.
Julieta Martinez, Joris Clement, Holger H. Hoos, James J. Little.
Revisiting additive quantization. In ECCV 2016 (26.6% acceptance rate)
Additive quantization (AQ) is a promising vector compression approach for large-scale approximate nearest neighbour search. We introduce an optimization method for AQ that pushes it beyond the state of the art.
Ankur Gupta, John He, Julieta Martinez, James J. Little and Robert J. Woodham.
Efficient video-based retrieval of human motion with flexible alignment. In WACV 2016
We formalize the problem of video-based mocap retrieval. We also investigate different retrieval methods for this task.
Julieta Martinez, Holger H. Hoos and James J. Little.
Stacked quantizers for compositional vector compression. In arxiv (2014)
Some of my early attempts to improve multi-codebook quantization. This approach is equivalent to enhanced RVQ, and has been superceeded by our work on revisiting AQ. The code is very accessible though!
Julieta Martinez, Holger H. Hoos and James J. Little. In 4th Workshop on Web-scale Vision and Social Media (VSM), at ECCV 2016.
Complement to our work on Revisiting AQ. Details our GPU implementation.
Frederick Tung, Julieta Martinez, Holger H. Hoos and James J. Little. In WACV 2015.
A vector is mapped to one of many hash functions, which improves accuracy at increased query time.