Matthew Trentacoste

Recovering academic. Distinguished raggamuffin.

I'm a scientist, engineer, photographer and motorcyclist — though not necessarily in that order.

I'm fascinated with the process by which our minds create the world we experience and how psychology and philosophy can impact our lives.

I'm a culinary data scientist. I spend my days studying how customers interact with the menu of a large on-demand restaurant. I think solving hard technical problems can help people eat less crap food.

Once upon a timeFind many more interesting facts in my CV., I was an academic researching human visual perception, including how image appearance changes with size and how perception of brightness & contrast influence the design of monitors. I have been an engineer, working at Adobe on photography apps such as Photoshop Fix, and at Dolby on the Vision line of high-end displays, amongst other things.

I'm a minimalist and a vagabond at heart. Some of my happiest times have been living out of a backpack. I crave movement.

San Francisco is where I'm most often found. Prior to there, home has been Vancouver, Pittsburgh, Washington DC and a few other places.

At other times I've been a cyclist, martial artist, powerlifter and health nut. I may be again some day. I'm a pretty damn good cook.

Select Projects

Recognizing Image Style

image stylesThe style of an image plays a significant role in how it is viewed. We predict the style of images, and perform a thorough evaluation of different image features for these tasks. Features learned in a multi-layer network generally perform best — even when trained for a different task.

Image Features in Python

image featuresThe quality of features can make or break a machine learning project. At PyData, I presented ways to extract features from images using Python tools that can be readily used by developers familiar classification and clustering algorithms. See also: slides and Github project.

Scale-dependent perception of images

chuck closeMy doctoral research examined ways in which perception of images changes when they are viewed at different sizes. Due to the organization of the visual system, the appearance of blurred and sharpened edges depends on the width of the edge. Besides the study, we propose several methods to address the change.

High dynamic range displays

DR37-PMy early academic career was dedicated to the development of HDR displays, possessing greater contrast and brightness than conventional monitors. Besides being of academic interest, I also played a role in the startup comercializing the technology, BrightSide Technologies, and later at Dolby.


Adobe Photoshop Fix
Recognizing Image Style
Image Features in Python (slides) (github)
Scale-Dependent Perception of Countershading: Enhancement or Artifact?
Color Splash
Manipulating Scale-Dependent Perception of Images (Ph.D.)
Glare Encoding of High Dynamic Range Images
Blur-Aware Image Downsizing
Defocus Techniques for Camera Dynamic Range Expansion
Dolby Vision display
Photometric Image Processing for High Dynamic Range Displays
Ldr2Hdr: On-the-fly Reverse Tone Mapping of Legacy Video and Photographs
Photometric Image Processing for High Dynamic Range Displays (M.Sc.)
High Dynamic Range Techniques in Graphics: from Acquisition to Display
Real Illumination from Virtual Environments (video)
Volume Rendering for High Dynamic Range Displays
High Dynamic Range Display Systems
Implementing Performance Numerical Libraries on Graphics Hardware (B.Sc.)

Find me

Get in touch via email (my initials at this domain), Twitter or Facebook.

What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information…