Rethinking the museum collection online.

Shelley Bernstein
Barnes Foundation
Published in
4 min readApr 11, 2017


In the coming months, we’ll be talking quite a bit about two projects currently in production — our website redesign and, running alongside it, a new presentation of the Barnes Foundation collection online. The collection project, in particular, is something we hope the industry will be interested in because we are heading in some different directions in both the visual presentation of the collection and the backend technologies employed to get there.

A collection online that mirrors the unique experience of the Barnes.

Albert Barnes wanted to use his collection to teach all audiences how to appreciate art, and he reasoned that grouping his artworks according to formal connections (rather than historical ones) made them more accessible. As a result, Barnes arranged his collection in “ensembles,” distinctive wall compositions organized according to formal principles of light, color, line, and space, rather than by chronology, nationality, style, or genre.

The ensembles were created and installed as a democratic ideal; audiences could correlate works visually through “light, line, color, and space.” Can an online presentation utilize computational image analysis to do the same thing?

Traditionally, museum collections have done the opposite and are often catered to the research audience, with rich semantic search and comprehensive metadata for every object, but almost no connections between works themselves. This works well for a sophisticated, goal-oriented user with a deep understanding of art and art history which, coincidently, describes most museum staff, but leaves most visitors without a way to explore the collection beyond searching for the names of artists that they’ve heard of or for words they like.

In rethinking the presentation of our collection online, we are taking a more experiential and Barnes-like approach. Just like our founder, we’re going to start by focusing on the needs and interests of non-specialist, non-professional audiences and, therefore, develop a collections website that makes the image primary. Thinking about Barnes and his teaching style very specifically, we will design an interface that highlights purely visual connections between artworks — colors, shapes, lines, for example — so a visitor always has a way to move forward and deeper into the collection in a self-directed fashion.

So, think of this as a next generation collection project. The standard catalog will be there, but we’ll be augmenting it with computational image analysis to cluster and filter lists of images by visual similarity. The goal will be to explore Dr. Barnes’ idea of “light, line, color, and space” while giving visitors a way to visually browse online in the same way he taught them to learn in his galleries.

Paul Cézanne. Bathers at Rest (Baigneurs au repos). 1876–1877. BF906. — What else is blue? What else is green? What else has vertical elements? What has matching qualities of light, space, volume?

Luckily, we have some models to follow for things like color analysis and we’ll be seeking a lot of help from our friends when it comes to light, line, and space — this is the fun part.

Refreshing our technical choices and open source as we go.

As most of our readers out there, I’ve lived this with you — the nightly dump of data out of TMS and into something more malleable, likely highly structured, and, hopefully using some set of standards (take your pick). Scripts and complexity abound. We started to ask ourselves if there was another way. Do we need all this complication? What’s different about today than yesterday? As part of this project I asked two leaders in our field — Aaron Straup Cope and David Newbury — seriously, if you could do it all over again what would you do? The challenge was not to think about how we’ve always done things, but to think way outside of that; what’s working in the world of code that hasn’t been applied to the collection online…yet?

The answer, in getting to fully start over, was to go for radical simplicity. Use the TMS API (seriously, words I never in a million years thought I’d be able to say) to extract data from the collections system and combine it with something more nimble on the other side like ElasticSearch. You can see Aaron’s proposed architecture and what we are headed toward, right here.

There’s more. There’s a whole lot more, but this is where I stop and this is where the team working on the project takes over. I’m extremely happy to say that Cassie Tarakajian, Sam Tarakajian, and Ana Giraldo-Wingler at Girlfriends Labs are building our new collection site. Cassie, Sam, and Ana will be blogging our path as we move forward and they are releasing code as they go. Aaron is continuing to consult for the duration and Micah Walter has been brought on board to help shepherd us to the finish line.

We are excited to share this project with our colleagues and hope the sector will find it useful; follow our Barnes publication to see these posts. We’ll be sharing a lot from now until our launch in late summer.

The Barnes Foundation collection online project is funded by the Knight Foundation and our code is open source. Follow the Barnes Foundation on Medium.



Shelley Bernstein
Barnes Foundation

Head of Product/CTO @ofbyfor_all. Digital consulting @the_barnes and others. Living in Far West Texas and loving it.