Computer Vision applications in Marketing
August 1
This week I will have a look at how new and emerging computer vision capabilities are being used in marketing. We all see these techniques in our everyday life. Next time you look at an ad or other marketing device, stop and think about the impact that computer vision has had in delivering the message specifically to you.
Much of the input for this week’s newsletter comes from Maria Yao’s Applied AI work on Topbots. If you want to delve deeper into this and other related issues Topbots is full of interesting papers.
Original Content Generation with GANs
We have spoken about GANs previously and we looked at how GANs can create manufactured images for fashion ads. Rather than going to the expense of hiring a studio, photographer and model, the GAN can generate hyper-realistic content. Pose-guided image generation can transform any image into different poses. This process uses a 2 stage generator (verses 1 stage in the first generation of GANs) and a discriminator. Here is an example from the Japanese company DataGrid.
Branded Object Recognition
Brands have to watch social media and other platforms to determine where prospects and customers engage with them. Image detection enables brands to see the threats and opportunities that exist. Gumgum is a social listening company that will identify brand logos and find good and bad reviews all over the web for their customers.
Product discovery through visual similarity
Many online shopping sites allow customers to use a filter function to discover new products. This requires an extensive use of tags which need to be manually assigned to products. Tagging is based upon the retailers perception of their product which may not match customer perception. Pinterest has an AI tool called Visual Search that allows visual product discovery, eliminating the need for tags. Consumers select an image they like and they will then be shown a range of similar items. This speeds up the consumers ability to find items that they like and will purchase.
Tracking Consumer attention and emotions
Face analysis algorithms are now powerful enough to assess consumer’s facial expressions and infer their emotions. Disney developed Facial Variational Autoencoders to determine how their audience responds to their films. Infrared cameras detect and capture audience reaction during movie screenings. This helps Disney understand what in their movies provokes certain emotions
Optimizing Conversion rates with Images
Yelp now curates beautiful photos for any establishment to maximize their conversion rates. Instead of using the number of likes (easily manipulated) to determine the best photos they judge photos on contrast, depth of field and alignment etc. Yelp then used a convolution neural network (I will examine neural networks in a future newsletter) to design a photo scoring model which guides them in photo selection.
Facial recognition for Personalized Customer Experience
A growing number of stores use facial recognition to identify valued customers as soon as they step inside the store. Sales clerks can see a customer’s purchasing history, preferences and taste profile. An AI powered analytic tool helps the sales person deliver personalized product recommendations.
Neutrogena has a skin scanning gadget that attaches to a smartphone which syncs to their Skin360 app. The app analyses your scans and determines your skin health and makes Neutrogena product recommendations.
Trying on Products Virtually
Sephora uses augmented reality and AI to create virtual shopping experiences. Customers try on the latest makeup through a mobile app. The AI will also help customers match skin tones to find the perfect product. A word of warning though, Sephora has recently been hacked. If you use Sephora, change your password asap. The more that we are online the more vulnerable we are to data breaches. More on this in coming weeks.
More and more computer vision and associated AI driven analytics are being used to try and influence customers to purchase. The use of these techniques is not only about reducing costs. It is also about increasing conversion to purchase. More and more, marketing is being personalized.
A bonus this week
We have talked about how voices are being artificially created and that there is potential for them to be used in various ways. Here is an example from Jolly Roger. They build a system to answer your phone when an unlisted number calls (usually a telemarketer or scammer). The system will recognize the call and then try and waste as much of the scammers time as possible. At the moment they use an AI to detect words from the caller and then respond with set recorded phrases however it is not difficult to envisage a future system that is more interactive. It is worth a listen.
Paying it Forward
If you have a start-up or know of a start-up that has a product, ready for market please let me know. I would be happy to have a look and give the start-up a shout out to my readers if it is something that I think they could use. If you have any questions or comments please email me via my website craigcarlyon.com
Till next week.