About me Who what how
I'm Matt. Or Trent. Or Matt Trent. I'm many places, but often found in San Francisco. I'm interested in nearly everything, most notably cooking, learning and creating new technology.
I'm a cross between a research scientist and a software developer. Practically, that means I do some combination of: observe and model phenomena, design algorithms based on said models, and craft usable implementations of those algorithms. My background is a mix of image processing, human visual perception and developing for new hardware and mobile platforms. I'm interested in machine learning, data analysis, sensors and how those will impact health and wellness.
Currently, I work at Adobe on mobile photography and data applications. Previously, I've worked with companies large and small, including Dolby, Pocket Pixels, BrightSide Techonologies and Mobify. If you're interested in what I can do, check out my curriculum vitae.
My Ph.D. thesis examined how our perception of images changes with their size and produced several methods to control how the image appears. I think it's pretty cool that machines can use knowledge of how we see to create better images for us.
I'm an avid biohacker and self-experimenter. I've benefited from taking the scientific approach out of the lab and applying it to everyday life. I feel it's not only important to collect data, but to distill it into actionable information. I was the Vancouver Quantified Self organizer, when I still lived there.
I really do like cooking.
I'm also an occasional world-traveller, photographer, martial artist, cyclist, miscreant, vagabond and wanna-be jalopy-racer. These and other activies have thus far occupied my time on this planet for the last 34.47 years.
Projects New or notable
The 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.
Effective machine learning combines the right algorithm with features that accurately capture the interesting aspects of the data. At PyData NYC 2013, I presented a number of ways to extract features from images using Python tools that can be readily used by developers familiar classification and clustering algorithms.
Top-selling selective desaturation photography for iPhone and iPad. Working with Pocket Pixels, I redesigned the app's graphics framework. Based on GPU acceleration, the latest version includes interactive image adjustments and faster performance. After release, the app rose to #6 in the US store.
My 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.
- Recognizing Image Style
- Scale-Dependent Perception of Countershading: Enhancement or Artifact?
- 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
- 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
- Volume Rendering for High Dynamic Range Displays
- High Dynamic Range Display Systems
- Implementing Performance Numerical Libraries on Graphics Hardware (B.Sc.)
Get in touch via email (my initials), Twitter or Facebook. Better yet, find me in person (that way we can have a drink). Whatever the reason: hiring me, collaboration, creating amazing things, tossing ideas around, or just plain old shooting the shit, I'm available.