You've now been added to our email list and will receive product and company updates from Axomic.
Published on 9th August 2018 at 13:42 by Alex Wetten
At OpenAsset, our product vision is to inspire people through visualization of the built world. As the leading Digital Asset Management provider for companies operating in the Architecture, Engineering and Construction (AEC), and Real Estate industries, we are continually offering improvements to our SaaS products that help to deliver this vision.
The main challenge within Digital Asset Management (DAM) is how to create useful metadata for images at scale. DAM users typically apply tags to files and projects using a keyword structure and a well-organised DAM allows users to easily search and compare images using these keywords. However, this is a manual process requiring an investment of time and money, which can be difficult to scale up as the business grows.
We believe that the introduction of Artificial Intelligence (AI) into our product could make the process of tagging images significantly more efficient, while also improving the quality of search. In order to launch AI-powered features within OpenAsset, our software engineers have undertaken a period of extensive research and development that continues to this day. Over the past 18 months, they have tested out a variety of approaches to establish the most valuable applications of AI for the AEC and Real Estate industries.
In this blog we shall outline some of the key areas in which AI can deliver improvements within OpenAsset, as well as some of the broader considerations for its application within Digital Asset Management.
In recent years, public interest in Artificial Intelligence has been steadily increasing and the concept has taken on a broad range of definitions. The term AI has come to encompass ideas ranging from automated production lines to self-conscious computer systems. As an emerging technology, the applications and implications of AI continue to be debated and companies often make grand claims for its potential that cannot be substantiated.
AI cannot run a marketing department or design the blueprint for a skyscraper and we have found that the best practice for implementing AI tech is to be clear about the strategic objectives that you wish it to serve. In so doing, AI’s specific capabilities can be adapted to suit your particular industry and the significant technical challenges presented by AI adoption can be more easily navigated.
AI is a subset of machine learning that deploys algorithms to large sets of data, enabling computers to mimic human intelligence. AI is central to the predictive modelling used by ecommerce or video streaming companies like ASOS and Netflix and anyone who receives targeted web ads will have encountered AI technology. Although AI is already embedded in web-enabled services we use on a day-to-day basis, the technology underpinning AI is rapidly accelerating and economic forecasters have identified it as a key area of growth that businesses will need to consider in order to remain competitive.
One of the main drivers for the rise of AI is the shift from local network storage to cloud computing. The increasing use by businesses of cloud-based networks has made available the large data sets necessary for machine learning. The ability of cloud servers to store nearly unlimited volumes of data and to deliver analytics in real time has greatly increased the ability of analysts to process and deliver insight from these data sets. By creating models of data that mirror the neural networks of the human brain, software engineers have been able develop the sophisticated algorithms that power AI.
The availability of AI services that can be incorporated into SaaS products via API has also encouraged their widespread adoption. The launch of services from companies including Google, Amazon, Microsoft and IBM has enabled a large number of businesses to benefit from their extensive research and development efforts without the prohibitive costs. This has led to a developmental ecosystem whereby larger companies undertake the sustaining innovation that improves the efficiency and powerfulness of AI tech, while smaller companies use disruptive innovation to open the technology up to new markets and users.
The other major technological advancement that has aided the adoption of AI technology is that of Natural Language Processing (NLP). This is a domain of computer science concerned with how to program computers to process and analyze large volumes of human (natural) language. This area of research is generally believed to have started in the 1950s with the eponymous tests of Alan Turing. However, AI technology helped to facilitate a recent breakthrough that has greatly enhanced the ability of computers to to grasp the nuances of natural language that are necessary for everyday interactions.
Services such as Google Translate previously relied on a technique called statistical language inference, whereby huge volumes of existing translations are cross-referenced to predict the meaning of individual words. With the introduction of neural networks, these programs could begin to understand how languages function by developing a complex model that mapped all the relations between the words in a language. It is this ability of AI services to understand and convey emotional intelligence that is of great value to companies deploying Martech services. Using AI-powered software, companies can better understand the needs of their customers and automated communications such as email programs and chatbots can be automatically tailored to suit individuals rather than groups.
Digital assets are any files that provide value to an organization. They may include photos, videos, PDFs and any other files that can be used to visualise your products and services. Storing these files on shared drives or networks can create problems, especially as a company grows or if they undergo a period of change, such as a merger or acquisition, or undertake a Digital Transformation project. Changes to naming conventions, data migrations or complicated folder structures often result in assets becoming lost, translating into a loss of potential revenue.
This is where a DAM platform like OpenAsset can help. By enabling users to efficiently tag files using a keyword structure, companies can ensure that any member of the organization can easily access and compare the best available visual assets. This is of particular value to AEC and Real Estate companies not only due to the large volume of visual assets they are typically required to store, but also due to the project-oriented nature of their work. As well as simplifying the process of searching for assets, OpenAsset offers integrations with the systems marketers use every day, including Deltek Vision, PowerPoint, InDesign and KA Synthesis.
The specific ways in which AEC and Real Estate firms may benefit from using a DAM vary across different industries. For Architecture firms, the ability to visually communicate design concepts is crucial to winning new business. Professional photographers are hired to take high quality images for promotional materials and ensuring that these assets remain searchable over time is essential to retaining their value.
Engineering and Construction firms are required to generate large volumes of materials for bids. Due to the high costs and slim margins of these projects, the bidding process can be fiercely competitive and the ability to assemble high quality project images at speed can help to distinguish a company and demonstrate their expertise.
Commercial and residential Real Estate companies operate in a fast-paced environment, with large teams collaborating on the production of marketing collateral. Admin users can set permissions for images, ensuring that all imagery is approved and reducing the risk of costly legal disputes that can arise from misused assets.
Below are some of the key benefits that a DAM platform can bring to AEC and Real Estate firms:
The development of our Image Similarity search feature introduced us to the opportunities of incorporating AI technology into OpenAsset. The feature uses data from the image recognition service Amazon Rekognition in order to calculate the probability of images featuring similar content. This will offer users a new way of discovering images that works alongside metadata-based searches. Users may already be familiar with image similarity or ‘reverse image search’ features in Google, Bing, Instagram or Pinterest.
Until now, users searching for images within OpenAsset relied upon keywords that had been selected by other users. Manually applying metadata in a large database of images is a time-consuming process, requiring staff or image librarians who are trained to use a specific keyword structure. Until metadata has been applied to these assets, they are not in a searchable state and are unable to deliver value to an organization. The Image Similarity search feature will offer an enhanced image search that offers value to users with no extra administrative effort.
When a customer selects an image in OpenAsset, the Image Similarity feature will return a selection of images from across the entire image library that it identifies as being visually similar. We believe that this feature should be easy and intuitive to use, offering value in a number of ways:
Users can discover images of buildings that share common architectural styles, similar patterns or compositions, or which feature specific building elements, such as an atrium or a spiral staircase
Searches can commence without the user having to decide upon specific search terms, increasing the opportunity for unexpected and creative image combinations
If valuable search terms are missing from the keyword structure, Image Similarity search could speed up the process of identifying images missing these terms
In the near future we are aiming to introduce within OpenAsset the ability for our software to offer AI-assisted keyword suggestions. Using image recognition it may be possible for OpenAsset to select relevant tags from your organization’s keyword structure, thus minimising the effort of image tagging, while maximising the effectiveness of search.
The main technical challenge in developing this feature is to be able to convert the image recognition data into valuable search terms that are relevant to the images used by our clients in the AEC and Real Estate industries. The models on which image recognition services are built are trained to recognise images to suit the broadest possible range of use cases. Their level of accuracy when analysing images of the built environment can vary and they often lack the specific contextual information that is of value to our clients. In order to understand the types of keyword suggestions that would be useful to customers, we closely examined the way in which our users classify their digital assets.
By analysing the data within OpenAsset, we discovered that users have amassed a total of around 15,000 project and file keywords. Leveraging our expertise of how customers within the AEC and Real Estate industries use keywords to tag their assets, we were able to develop a broader picture of how this data could be used to improve search functionality. With more focussed analysis, we established patterns in how users applied certain keywords, arriving at a smaller set of a few hundred search terms that frequently recur across OpenAsset.
Developing a set of useful keywords was the starting point of our AEC and Real Estate keyword ontology. The keywords function like a vocabulary that can be used to describe images. In order for our software to predict the contents of new unseen images, it requires a set of rules that govern how the keywords relate to these images. This is known as a classifier, which is an advanced system of algorithms that can translate the generic output of an image recognition programme into meaningful keyword suggestions. Using our analysis of how users select keywords, we are aiming to develop a classifier that can respond to the particular requirements of images of the built environment.
OpenAsset users will soon receive inline suggestions for project level keywords and further down the line, file level keywords as well. As AI technology advances, it may become possible to fully automate the process of tagging assets. By gradually incorporating greater levels of AI- and IR-powered technology, we are ensuring that our product is delivering tangible benefits to users while setting ourselves up to take advantage of future developments.