 Kiwana Tato, I'm Jennifer Taylor-Moore, Curator of Collections at the Sargent Gallery Te Wariwarehua Wanganui. And I'm Paul Rowe, the CEO of Vernon Systems, and we're here to tell you about how we made the most out of our online collection project with limited resources. So the Sargent Gallery has a small team of about 11 staff with no in-house IT or design staff. The Gallery Building, which opened in 1919, is a category one listed heritage building. The Sargent Gallery is in the midst of a major $35 million redevelopment programme, which includes seismic strengthening of this iconic century-old building and construction of a new wing to allow for greater exhibition space, a far larger collection storage area, an education area and dedicated spaces for events. The Sargent Gallery's nationally significant collection comprises of over 8,000 artworks and archival items, spanning four centuries of European and New Zealand art history. It is one of the country's most comprehensive surveys of New Zealand art history and a regional centre. Following the Canterbury earthquakes, the Gallery's heritage building was identified as an earthquake-prone building and was closed to the public in 2014. We relocated to a smaller temporary site called Sargent on the Key to maintain public programmes during fundraising for earthquake strengthening of the heritage building. The exhibition space is significantly more restricted at our temporary premises and this lessens opportunities for us to display the collection. As a result, there was an urgent need for online access to the collection during this redevelopment period. We set a goal of launching a website for the online collection within four months of the start of the project and with a budget of $23,000. We were keen to build a website that provided open access to the collection and offered a variety of options for visitors to explore the collection online. A particular focus was to make the site easily accessible for online visitors who had no prior knowledge of our collection. The Gallery's operating budget is very modest and any large funds that might have been available for this project had already been channeled towards the fundraising for the redevelopment. The budget constraints meant that there were not enough resources to spend hours retrospectively inputting data and keywords into the collection records. So we had to look for ways to quickly clean up the data and prioritise gaining copyright clearance with the artists in order to display their images on the site. Our collection was already cataloged in Vernon Systems' Collection Management Database and Vernon Systems offered to work in partnership with us to develop the new site. This site was built on top of an existing online collection platform built by Vernon Systems and this allowed us to make better use of our budget focusing on improving the catalogue data, adding new innovative features and incorporating our gallery branding. We made the decision to put the entire collection online including records where the image was not yet available. The full collection provided many more records to connect across the collection and online access is now helping us to spot data that can be improved. The online access is also helping us reach copyright holders who we weren't previously able to contact. The images have been put online even if they're not perfect as in this example where the image will be cropped when we have the time to do so. We've allowed reuse of images wherever possible and clearly mark images for which there are no known copyright restrictions. We made it easy to find images that are available for reuse and search results to be filtered by the current rights for each artwork and here you can see our main rights categories. The site uses responsive templates to provide layouts for mobile tablet and desktop screens. Visitors and staff can easily look up the details of an artwork from any device. On smaller screens we automatically display smaller image sizes to improve the speed over slower connections. We would first give a quick thank you to Shanoi Sam, our developer at Vernon Systems who built this website. The first thing we did is we used John Fawkes existing research on typical museum visitor types and we used that as a way of assessing the potential features within the website and seeing how they might fit the people we were trying to target. For example, we've ended up with a home page where it provides a number of simple links into the collection by things like colour, object types, but we've also considered adding in an advanced search page so that we could still meet the needs of researcher types so people that knew in advance exactly what they were looking for. We also made a commitment to using computer vision so tools where we could automatically analyse the images and the artwork image. So a couple of the different tools that we made use of were Google Cloud Vision and Color Thief. We also looked at what we could do quickly with the limited staff time that we had available. So for example we just spent an hour adding nationalities for the 40 countries of birth that were listed in the Artist Database and that means now that for almost all of the artists in the collection of the search filters. So for example if the country was Japan for the country of birth, Japanese appears as one of the nationality options. Most of the catalogue records didn't have descriptions in the original database so we looked at ways that we could generate a natural sounding description by combining the object type, the production place and the production date. So it automatically does a look up to see how many other works there are online of the same type to give the visitor a sense of scale for the online collection. And it also embeds links to allow you to easily jump to all the things that are made in the same place or are of the same object type. And so here we see a typical example. This is one of 1800 photographs in our collection. It was made in Whanganui in 1907. We also spent a couple of hours adding links for the most prominent artists in the collection to Wikipedia and to Te Ara, the Encyclopedia of New Zealand. So that provided a way for almost a third of the whole collection to connect through to these additional resources. So we thought it was essential, absolutely essential that the website didn't exist in isolation from the rest of the web. Similarly, we looked at pages that already existed in Wikipedia and if the gallery had significant holdings of that artist, we've placed links back from Wikipedia into the website and those links either go to an artist page or a specific artwork if it's a particularly famous work. The gallery's also looking at adding in Wikipedia pages where they're missing for artists that they hold in their collection. So we made the decision to build the website as a standalone online collection website apart from the main gallery content management system and that allowed us to build on an existing working online collection product and not have to try and rebuild those features within the main content management system. It also let us focus on a layout that best suited the online collection. So for example using the full width of the web browser to display the content piece of a whole range of different features that we couldn't have built within the main website and so that allowed us to do more within the budget that we had. The online website's also less likely to be affected by changes in the design in the main website over the longer term. Prototyping was an essential step for us to assess potential features that we had look at how we might build those and see whether they were actually viable or not. Here we see an early example where we used our product but just as an unbranded wireframe essentially to explore the data and see how it might work on the website. In an early design we showed the dominant colours of each artwork in round circles to evoke an artist's palette but through our testing we found that not all visitors realised that those were things they could click on to navigate to other artworks so that was changed later on in the design. Even in the smallest projects you'll benefit from user testing. We made use of staff at Sargent Gallery and at Vernon Systems that weren't otherwise involved in the project and we asked for volunteers from outside of our organisation and so those testers covered a range of ages and interests and they helped us find problems on the website and suggest potential improvements. So one of the things we were encouraged to do was to use planar language within the design. So for instance on the rights page where you can see what rights apply to a specific image we've introduced an open and closed padlock symbol to make it easy to see which images are actually available for reuse. The website provides a number of different ways to explore the collection such as browsing by colour, looking up the highlights of the collection going by object types but not everybody realised those options were there so we added shortcuts at the top of the home page and new menu options so that those options were always above the fold. We also had to make use of the data that we had not the data that we wished we had and so one example is that the object type field in the database was fairly consistent and so we've provided that as one of the entry points on the home page but we also did an automated look up again to see how many works fell into each of those categories so immediately on the home page you get a sense of scale as to how many items there are so we can easily see that there's 1800 drawings within the online collection. We also made use of that data on the collection detail page to ensure that there were no dead ends so provide links to easily jump from one artwork to other related artworks. We wanted people to linger on the site and browse and make serendipitous discoveries we also looked at ways that features could serve more than one purpose so on the site there's a directory of all the artists that are represented within the Sergeant Gallery's collection but that page also provides a way for search engines to get to every single artist page and every single artwork page from there. Lastly we looked at what we could do with automated image analysis so one easy thing we could do was pull out the orientation of the image was it landscape, portrait or square and provide that as one of the filters on the search page so here for example are all of the artworks by Gretchen Orbrach that are in landscape orientation and then we made use of Google Cloud Vision to automatically add subject tags to the artworks now computer vision is a complex field and the results aren't always perfect and so we had to develop strategies for dealing with that. We've added administrator tools where the staff at Sergeant Gallery can remove a tag that isn't appropriate for an artwork or can remove a tag across the whole website so one example is that Google Cloud Vision tagged almost all of the photographs as stock photography and that wouldn't have gone down particularly well with the artists and so that's an example of a tag that got removed from the whole website but even though those tags aren't always perfect they do provide some interesting results and new connections so these are a bunch of works that have been tagged as circle none of them are strictly circles but they are providing a new connection between those works because they've got similar shapes and lines. We used an open source tool called colour thief to extract what the most dominant colours were from the original 16 million colour palette that's in the image however unless you've got a really enormous collection hardly any of the works will share the same precise colour and so what we had to do was map those colours down to a smaller palette and we used the official named web colours 140 colours and because of that much more of the objects now share the same matching colours the only disadvantage is that the colours are moving slightly away from the original precise image colours because we used named colours we could also index the words for those colours and so we index the precise colour name which might be Crimson for example and the colour group such as red and that means people can now include a word for a colour as part of a tech search and they'll find the works even though those words weren't in the original catalogue record we also looked at complying with accessibility guidelines so making sure that the website worked on a number of different device sizes adding alternative texts for images allowing people to zoom within the web browser to size the content to a size that made sense for them accessibility compliance is important because it makes sure that your content is usable by the widest range of people as possible through the prototype there are a number of ideas which we decided not to keep so I was just going to give a couple of examples of those we did originally look at changing the background colour of the artwork page to match the artwork so we already knew the colour through our colour analysis and here's a typical result however we found that for some artworks that background colour just became distracting and it was impossible to write generic enough code to avoid that being a problem in some cases we also considered making a diagram that showed how the colours in the artworks changed over the decades but what we hadn't considered is that almost all artworks included some common neutral colours and that's the colours that ended up bubbling up to the top so we ended up with a graph which was essentially 50 shades of beige and an idea which we ended up keeping to come back to later to see if there was a way we could implement that better and lastly we also looked at whether we could generate full sentence captions for the images now I think this is the hardest part for computer analysis and it's where we saw the most variable results we looked at a test website that a company called COGAP had put together using the National Museum of Sweden's photo or painting collection and from that we decided that it just wasn't really good enough for us to use in the project at this stage with the finished website we've embraced the concept of create once publish everywhere and we're using the online collection to engage the public in different ways such as this example where we shared via social media feed a set of works featuring mistaches as part of the Movember event and the tag which was being searched on as facial hair we've also included simple options for visitors to share a page on common social media platforms and this is done using the open graph metadata standard which ensures the correct title, description and image from the page are carried over into the user's social media post one of the most exciting outcomes has been using the new site to engage with the collection in different ways than were possible previously particularly using the highlight section of the website and we've developed several initiatives as a result that involve digital as well as linked exhibitions the collection focus series highlights a group of works by a featured artist from the collection online but also as a physical exhibition in the gallery the works currently on display highlight feature allows people to see at a glance which collection works are available to view in the gallery at any moment and also provides access to additional information beyond the traditional exhibition labels and we have a kiosk in the gallery where people can do that next year the gallery will launch a new initiative called my choice and this gives the public opportunities to make their own selections of works for display in a virtual exhibition online the person making the selection will also add their own personal comments about why they selected these works this series will be launched in May next year with a physical exhibition of our own choice he can't wait followed by online selections by members of the community which will upload on a monthly basis we're also wanting to make available to the general public the chance to submit their own selections via social media and will select a few of those to also feature on the site we're now planning what we can do next and Google analytics is a valuable tool for assessing the use of the website we can see that we're reaching a much wider audience with a third of visits coming from outside of New Zealand Wikipedia has proven to be a key website and helping people to discover our collection and we're continuing to work steadily through requesting copyright permissions from artists and copyright holders to increase the number of images that can be displayed almost every artist approach so far has responded positively we've had some ideas that didn't make it pass a prototype stage such as presenting more visual displays of the breakdown of the collection and this is a concept we hope to look at further in another development phase we're poised to launch a new feature on the website called Sergeant Mosaic this provides an overview of all the collection images on one screen sorted by either the production date or the artist's surname is pictured here in a distinctive style such as Gretchen Albrecht's ovals in the third row and this kind of view provides yet another way for visitors to zero in on a work that interests them by clicking on a small thumbnail it brings up a larger preview and then takes them to the detail page if they want to look further early next year this facility will be installed in the gallery on a large 65 inch touchscreen there's been very positive feedback from artists and visitors and we're excited to win the NDF 2017 Best New Digital Exhibition Collection Award we're also invited to speak about the project at the North American Museums and the Web Conference in Vancouver in April and we're also named runner up for the Best Digital Service Award at the recent Association of Local Government Information Managers Spring Conference the online collection has generated new interest in the gallery people have discovered works they did not know were in our collection members of public have contacted us with information about artworks on the site and as a result we've gained new knowledge about our collection the site has also been extremely helpful when visitors arrive at the gallery expecting to see a specific work on display right at that moment that they visit and usually more often than not it is not on display when they want so our staff are able to direct them to the online terminal on the gallery and they leave much happier we've focused our staff time on a few areas where data could be improved to the maximum with minimal effort and on the larger task of rights clearance which provides a better quality visual experience for visitors by carefully selecting and testing potential features we've completed the project on time and within budget computer vision has made it easier to discover the images and has opened up new connections within the collection we've connected with a far wider audience than would have been possible previously and the public are more actively engaging with the collection and the information that it contains thank you for your time just one question for me obviously you got the shout out from digital New Zealand but other than digital New Zealand where else are you finding your images appearing now almost all of the traffic comes through Google search so they're appearing everywhere and through Google analytics we can see those visitors are just coming from all over the world there is a large amount of traffic coming from Wikipedia as Jennifer noted so that's made it more obvious to try and make better connections there, there's still probably more artist pages that we could be linking through there to try and create a better network I guess there sorry you're getting a leachman one too here have you ever thought about allowing the public to tag your art on your website did you consider that that is something I think our main problem is the resources that we would have to manage that and at the moment our resources are extremely limited like I'm doing this alongside everything else and there's no other people in the team you'd have to curate it and make sure that the tags weren't inappropriate and I think that will be a problem but definitely it will be something that we'll be interested in doing but just not at the moment okay thank you this is a really simple one are the links to Tiara and Wikipedia in Vernon or on the website they're on the website and not in Vernon they're in the person record so where the artists details are they're in Vernon just as an external file link and marking whether it's public or not thank you you offer a residency to is it two artists a year are you going to post their work on there as well or is it yeah that's one of the highlights that I want to build which is all about the tidy residency and the artists have been featured on there but this site is very much focused on the collection and at the end of every residency we always acquire an artwork from the artists so I would like to build one of that definitely there's endless scope for topics that we could cover in those highlights so yeah it's just a matter of building up the I want to have like a brief paragraph interpretive description for each work in it and things so yeah it's just about the ideal in managing it with our limited time and resources so yeah thank you