How Computer Vision is making us more effective at our jobs
Computers are quickly becoming more able to understand the world around them — recognising faces and shapes — and making connections between the signals and data they receive.
This is changing the world around us, unlocking new and exciting possibilities for consumers and businesses.
Computer vision is a branch of artificial intelligence (AI) that involves teaching computers how to recognise and discern between objects in the real world by analysing the different components (e.g. the edges, textures or colours) in an image.
As human beings, we use our eyes and brains to analyse our visual surroundings without question. A computer, on the other hand, cannot do that automatically. It needs algorithms and data to learn what it’s “seeing”. Once it gets it right, however, in many situations it can do it unbelievably well.
As improvements to AI continue, consumers are beginning to feel its presence more prominently in their day-to-day lives. When you tag someone in a photo on social media, more often than not the person’s face tags automatically through data recognition. Or when you sync up a form of payment to certain apps, you can simply take a picture of the card you want to use and the numbers will ‘auto-fill’. And how many of us have used the filters available on Instagram and Snapchat to edit our faces for a selfie?
Even the trusted password and PIN are being replaced by biometric security, using your face to unlock a computer or phone through advanced facial and audio recognition.
Using this technology for the better
Given the growing prominence of computer vision in daily life, the onus is on businesses to think about how they can best use it to improve the customer experience.
One way of approaching this challenge is to focus on improving the efficiency and quality of processes. This allows companies to build expertise and experience in developing tools for back-end processes that could eventually be converted for use amongst consumers.
Stock inventory management woes
Working in the luxury fashion retail industry, one of the biggest challenges we and our peers face is stock inventory management.
It might not sound that important, but stock inventory management can cost retailers millions in both money and wasted hours a year.
For many retailers, especially those with an extensive catalogue, it is relatively common to find duplicate entries for the same item. This can be down to countless factors — perhaps the stock was marked incorrectly when it arrived, or maybe it was dispatched to a different warehouse which had not handled that inventory previously. Whatever the reason, the effect is the same — duplicate entries in the catalogue creating a clunky customer experience. And, significantly, underneath those duplicate entries is a mountain of work processing the items to get them into the catalogue.
This is where computer vision comes in. By using innovative techniques in deep learning and segmentation, we can now use a photo to identify if an item has previously been entered into the catalogue.
This isn’t quite as straightforward as it sounds, but the technology we are experimenting with allow us to take a photo and immediately receive feedback as to whether it is already in the catalogue.
We are working on perfecting the system, with the ultimate goal of preventing us from duplicating the entire inbound inventory process, which can include taking photos and styling alongside other merchandise in a studio, writing product descriptions and more.
Streamlining the catalogue process
Catalogues contain millions of studio quality images, one for each item. Yet for the same individual item, it is possible to have many different versions of the photograph depending on the light its captured in, creases, folds, size and even the way the item is placed on to the mannequin when the photo is taken.
Another challenge is that there will be different items out there that look extremely similar to the one in question, especially if the factors of light, folds etc. fall in the same way.
This means when you need to retrieve a matching item from millions of items online using a mobile-captured image of the item in the warehouse (which no doubt looks different), you encounter problems.
Our partnership with Microsoft
Knowing the benefits of computer vision, we have invested time in reaching an important AI milestone which will deliver a less wasteful inventory management process and allow our employees to work on more important tasks.
We’ve been working with Microsoft to test a newly developed computer vision technology, based on deep learning and segmentation (including Otsu thresholding, Grabcut and Tiramisu Semantic Segmentation), which can accurately spot duplicates in the catalogue through separating the image from its background and efficiently comparing them with the millions of items already available in our catalogue. This will save staff precious time sending products to a photo studio when they are already in our system. By using machine learning functions to transform pictures, the system analyses an uploaded photo and immediately see if it’s already in the catalogue — voila!
While freeing up staff to focus on more important work, there are other uses for computer vision too. Customers could also upload a photo to see similar products and monitor trends on websites like Instagram and Facebook, redefining the way customers search the YNAP catalogue while also giving us a heads up on the images that get the most engagement.
The long-term objective is to make the whole digital production process more automated and self-learning, step by step, which will allow us to move human work towards higher-value tasks and quality enhancements, rather than data entry. This partnership with Microsoft is a huge stepping stone in making that objective become a reality.