A bright minded start-up, Oodls, were looking to put their pioneering business model into action. Based in Edinburgh, Scotland, the company’s aim was to allow Instagram users to sell their images directly to businesses. The company intended to facilitate this process, allowing for convenient, easy, and affordable transactions, with mutual benefits for both parties. This forward-thinking business model promised to disrupt the stock photo industry.
To progress this business plan, Oodls required a web-based portal to provide the functionality. Utilising cutting edge software, the system would integrate, categorise, and present a wide assortment of photos from Instagram. These images would then be available for commercial designers to browse and interact with in a user-friendly format. Oodls intended to take care of all licensing and copyrighting, as well as paying the original owner, through this new platform.
With this project it was important that search results were relevant and comprehensive, ensuring that buyers could find exactly the image they were looking for. This posed an interesting development challenge given that Instagram’s simple tag system for content searching was incredibly restrictive. Many images would be tagged inaccurately or incompletely, limiting any system scouring the site for relevant content. This obstacle had to be overcome for Oodls’ business model to succeed.
The process that this solution sought to automate was a complex one. Categorising images without simple recognisable datasets would prove incredibly challenging for a software system. In computing, complex problem solving is best achieved with cognitive technology. The challenges presented by this project called upon such a solution. Artificial intelligence tools such as machine learning can be used to achieve near human levels of insight. A.I. offered the best route to categorising images from Instagram without reliable tags pre-emptively in place.
We applied these techniques and sciences to the project, and the result was an advanced search facility. The technology was powered by deep neural networks, based on Microsoft Conative Services. As a Microsoft silver partner, Shoothill have access to cutting edge technology, and the expertise to leverage it. Using a large sample of training data, we built a machine learning model from the ground up. This model was taught to interrogate any image and appraise its content.
A system operating machine learning must first learn how to interpret data. During this training phase the software is provided a large sample set and told how each entry should be interpreted. The system then uses pattern recognition across to construct mathematical algorithms which will inform its decision making moving forward. By applying this algorithm to future data, software can refer to the associations it has learned for intelligent analysis.
Our system developed for Oodls was trained with countless pre tagged images. These allowed it to determine what it should look for in a photo to assign each tag. As a result, the A.I had a strategy to identify characteristics when presented images from Instagram. By drawing on its own self written algorithms, it was successfully able to determine what tags to assign any given image. Beyond being entirely autonomous, this task could be performed far swifter by the software than any human end user could achieve.
The final cognitive model was able to curate and catalogue a vast and growing library of images through the Instagram API. Interested parties can access the portal and browse an array of images available for purchase. The A.I. allows the end-user to be incredibly specific with their search terms, such as “an old man with a black dog”, and quickly finding exactly what they are looking for. The system returns any results in which the machine learning mind identified both an old man and a black dog, based on its experience finding similar artefacts during the training phase.
The site also included several useful filters, so that a potential buyer can find images based on several criteria. Values such as date, author, dimensions, image quality, orientation, or the number of people in any given photo. All of these are automatically determined thanks to the Shoothill A.I. Furthermore, the system continues to curate images from Instagram, many of which are already tagged. These existing tags are used to expand on its understanding of images, making future searches more accurate. The A.I. then adds to them, to create a comprehensive meta-description. This ensures a user is provided a complete and relevant selection of images, without missing out on potential options due to obscurity.
This software proved game changing. The versatile use of A.I. in the project had the potential to disrupt the industry. This unique solution enabled Oodls innovative business model, allowing their customers to find online imagery ideally suited to their needs. Cutting edge technology was able to take a simple yet challenging idea and make it a reality. The sleek technology used to achieve this end result proved a major boost to the brand value of the digitally native customer.
Both parties saw this project as a massive success. Oodls were thrilled with the result, which boosted their brand image and provided the means to do business the way they had dreamed, while remaining efficient and effective. We at Shoothill were proud of the skill we demonstrated throughout the complex development. Artificial Intelligence was, and remains, a challenging tool to utilise, but we were able to create a first-rate system in a narrow time frame.
The work on this project taught our team much operationalising a functioning cognitive computing model. This led us to create the Photo-vamp suite, another Shoothill A.I. innovation which can automatically augment images. Photovamp is able to identify elements of a photo, caption them, and even colourise black and white photos in a smart and intuitive way at the click of a button. For example, the software is able to read archived photos of WW2 planes, determine the model of craft and recolour to bring history to life.
This project offers a tiny glimpse into what is possible with A.I. both now and in the future, and the incredible transformations to any number businesses we can offer our customers.