• LinkedIn • GitHub •  jazs.pro[at]gmail.com

Since April 2024, I work at Meta as a Research Scientist, applying advanced ML/AI methods to prevent harm on the platform. Prior to this, starting in March 2021, I’ve been exploring applying machine learning to stock selection, and building open source tools for ML/AI ( soleil, ploteries, extensible). Before that, starting in 2017, I worked at various projects at AWS/Amazon, including Amazon Scout, Rekognition and Textract. In my last role at AWS, I led the science team for Amazon Lookout for Vision through its general availability launch. Before that, I was a researcher at Technicolor in Rennes, France, for 6 years. I completed a PhD at INRIA in Rennes, France, and a B. Eng. and M. Eng. at McGill University, in Montreal, Canada.

Patents

• US11481683 Machine learning models for direct homography regression for image rectification
• US11741592 Generation of training data from a single image
• US10999607 Methods, systems and apparatus for electro-optical and opto-electrical conversion of images and video
• US9244948 Computer tool with sparse representation
• US12147878 Feedback-based training for anomaly detection
• US11184581 Method and apparatus for creating, distributing and dynamically reproducing room illumination effects

Publications

Learning a Complete Image Indexing Pipeline

Himalaya Jain, Joaquin Zepeda, Patrick Pérez and Rémi Gribonval.

In Computer Vision and Pattern Recognition (CVPR), 2018.

SuBiC: Supervised structured binary codes for image search (Spotlight)

Himalaya Jain, Joaquin Zepeda, Patrick Pérez and Rémi Gribonval.

In International Conference on Computer Vision (ICCV), 2017.

Kernel Square-Loss Exemplar Machines For Image Retrieval

Rafael Sampaio de Rezende, Joaquin Zepeda, Jean Ponce, Francis Bach, and Patrick Pérez.

In Computer Vision and Pattern Recognition (CVPR), 2017.

Maximum Margin Linear Classifiers in Unions of Subspaces

Xinrui Lyu, Joaquin Zepeda, and Patrick Pérez.

In British Machine Vision Conference (BMVC), 2016.

SPLeaP: Soft Pooling of Learned Parts for Image Classification

Praveen Kulkarni, Frédéric Jurie, Joaquin Zepeda, Patrick Pérez, and Louis Chevallier.

In European Conference on Computer Vision (ECCV), 2016.

Approximate search with quantized sparse representations

Himalaya Jain, Patrick Pérez, Remi Gribonval, Joaquin Zepeda, and Herve Jégou.

In European Conference on Computer Vision (ECCV), 2016.

The CNN News Footage Dataset: Enabling Supervision in Image Retrieval

Cagdas Bilen, Joaquin Zepeda, and Patrick Pérez.

In European Signal Processing Conference (EUSIPCO), 2016.

Supervised Learning Of Low-Rank Transforms For Image Retrieval

Cagdas Bilen, Joaquin Zepeda, and Patrick Pérez.

In IEEE International Conference on Image Processing (ICIP), 2016.

Exemplar SVMs as Visual Feature Encoders

Joaquin Zepeda and Patrick Pérez.

In Computer Vision and Pattern Recognition (CVPR), 2015.

Block Prediction Using Approximate Template Matching (Oral)

Joaquin Zepeda, Turkan Mehmet, and Dominique Thoreau.

In European Signal Processing Conference (EUSIPCO), 2015.

Max-Margin, Single-Layer Adaptation of Transferred Image Features

Praveen Kulkarni, Joaquin Zepeda, Frédéric Jurie, Patrick Perez, and Louis Chevallier.

In BigVision Workshop, Computer Vision and Pattern Recognition (CVPR), 2015.

Learning the Structure of Deep Architectures via l-1 Penalization

Praveen Kulkarni, Joaquin Zepeda, Frederic Jurie, Patrick Pérez, and Louis Chevallier.

In British Machine Vision Conference (BMVC), 2015.

Hybrid Multi-Layer Deep CNN / Aggregator Feature for Image Classification

Praveen Kulkarni, Joaquin Zepeda, Frederic Jurie, Patrick Perez, and Louis Chevallier.

In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2015.

Pose and expression-coherent face recovery in the wild (Oral)

Xavier Burgos, Joaquin Zepeda, Fran ̧cois Le Clerc, and Patrick Pérez.

In International Conference on Computer Vision (ICCV) Workshops, Robust Subspace Learning and Computer Vision, 2015.

Learning Sparsity Inducing Analysis Operators for Discriminative Similarity Metrics

Cagdas Bilen, Joaquin Zepeda, and Patrick Pérez.

In Signal Processing with Adaptive Sparse Structured Representations (SPARS), 2015.

Feature Learning for the Image Retrieval Task

Aakanksha Rana, Joaquin Zepeda, and Patrick Perez.

In Asian Computer Vision and Pattern Recognition (ACCV) Workshops, 2014.

Transfer Learning via Attributes for Improved On-the-fly Classification

Praveen Kulkarni, Gaurav Sharma, Joaquin Zepeda, and Louis Chevallier.

In IEEE Winter Conference on Applications of Computer Vision (WACV), 2014.

The Stanford / Technicolor / Fraunhofer HHI Video Semantic Indexing System

André F. de Araujo, Fernando Silveira, Haricharan Lakshman, Joaquin Zepeda, Anmol Sheth, Patrick Pérez, and Bernd Girod.

In TRECVID, 2013.

Image compression using the Iteration-Tuned and Aligned Dictionary (Oral)

Joaquin Zepeda, Christine Guillemot, and Ewa Kijak.

In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 793–796, 2011.

Image Compression Using Sparse Representations and the Iteration-Tuned and Aligned Dictionary (Journal)

Joaquin Zepeda, Christine Guillemot, and Ewa Kijak.

IEEE Journal of Selected Topics in Signal Processing, 2011.

The Iteration-Tuned Dictionary for Sparse Representations (Oral, 2nd best paper)

Joaquin Zepeda, Christine Guillemot, and Ewa Kijak.

In IEEE Workshop on Multimedia Signal Processing, 2010.

Approximate nearest neighbors using sparse representations

Joaquin Zepeda, Christine Guillemot, and Ewa Kijak.

In International Conference on Acoustics, Speech and Signal Processing, pages 2370–2373, 2010.

New sparse representation methods; application to
image compression and indexing (PhD Thesis)

Joaquin Zepeda.

PhD thesis, Université de Rennes 1 (INRIA), 2010.

SIFT-based local image description using sparse representations

Joaquin Zepeda, Ewa Kijak, and Christine Guillemot.

In IEEE International Workshop on Multimedia Signal Processing, 2009.

Tandem Filterbank / DFT Code for Bursty Erasure Correction

Joaquin Zepeda.

Master’s thesis, McGill University, 2006.