Check out my publications and our shiny 16-GPU Monster-Class Supercomputer.

High-Throughput Approach to Biologically-Inspired Vision in the Real-World

Summary: The overarching goal of this project is to dramatically accelerate the development of computational theories of how the visual cortex accomplishes object recognition. In addition to advancing our understanding of how the brain works by generating new, experimentally-testable, hypotheses his approach also holds great promise for the development of new artificial vision systems. A key innovation in this work is the ability to leverage the computational power of cutting-edge technologies (e.g. GPGPU and Cloud Computing) to provide new insights into this fundamental problem.

  • Pinto N, Doukhan D, Dicarlo JJ, Cox DD - A High-Throughput Screening Approach to Discovering Good Forms of Biologically-Inspired Visual Representation. (PLoS 2009)
  • Cox DD, Pinto N, Doukhan D, Corda B, DiCarlo JJ - A High-Throughput Screening Approach to Discovering Good Forms of Visual Representation (COSYNE 2008)
  • Pinto N - Visual Object Recognition in the Real-World: A Biologically-Inspired, High-Throughput Approach. (M.S. Thesis 2007)
  • Pinto N - Biologically-Inspired Real-Time Frontal Face Detection on Playstation 3s. (MIT/6.866 2007)

Establishing Good Benchmarks and Baselines in Artificial and Biological Vision

Summary: Progress in understanding the brain mechanisms underlying vision requires the construction of computational models that not only emulate the brain's anatomy and physiology, but ultimately match its performance on visual tasks. In recent years, "natural" images have become popular in the study of vision, and have been used to show apparently impressive progress in building such models. In this work, we demonstrate that tests based on uncontrolled natural images can be seriously misleading, potentially guiding progress in the wrong direction. Instead, we re-examine what it means for images to be natural and argue for a renewed focus on the core problem of object recognition -- real-world image variation.

  • Pinto N, Majaj NJ, Barhomi Y, Solomon EA, Cox DD, DiCarlo JJ - Human versus machine: comparing visual object recognition systems on a level playing field. (Submitted)
  • Pinto N, Barhomi Y, Cox DD, Dicarlo JJ - A systematic comparison of state-of-the-art visual representations for solving invariant object recognition problems. (Submitted)
  • Pinto N, Dicarlo JJ, Cox DD - How far can you get with a modern face recognition test set using only simple features? (CVPR 2009)
  • Pinto N, Dicarlo JJ, Cox DD - Establishing Good Benchmarks and Baselines for Face Recognition. (ECCV 2008)
  • Pinto N, Cox DD, DiCarlo JJ - Why is Real-World Visual Object Recognition Hard? (PLoS 2008)
  • Pinto N, Cox DD, Corda B, Doukhan D, DiCarlo JJ - Why is Real-World Object Recognition Hard? - Establishing Honest Benchmarks and Baselines for Object Recognition. (COSYNE 2008)

Understanding Endogenous and Exogenous Factors during Early Visual Development

Summary: Within a few weeks after birth, and possibly sooner, a human infant begins to exhibit some sophisticated visual skills. This unsupervised visual learning is known to be a very computationally complex task. Understanding how the infant visual system manages to overcome this challenge is of great basic and applied significance, with implications for neurological disorders like autism on the one hand and the design of autonomous robots on the other. In this work, we propose to address this fundamental question by identifying endogenous and exogenous factors that facilitate learning by a nascent visual system, and we develop simple plausibility proofs through computational simulations.

  • Pinto N, Moulson MC, Sinha P - The Benefits of Poor Acuity for Face Learning. (VSS 2009)
  • Pinto N, Moulson MC, Sinha P - Discovering Faces in Infancy. (In Prep)