The dynamic development of data processing techniques, constantly growing computing power and increasingly effective algorithms allow the use of graphical methods of information search - the technique called reverse image search. It is available today through many web applications and services like Google search engine or other solutions, such as the TinEye website or Magento platform.
Most users, when searching for any information, websites, etc. simply enter queries in plain text in the search field - mainly the key words, product names, sentences, fragments of texts. This way of using a search engine, especially the most popular one, branded by Google, is generally well known to every Internet user. However, in a situation where, for example, we want to find a specific model of a shoe, but the only information we have is a photo of a given clothing, entering any text phrase is not an option.
In this case, the solution is to use the so-called reverse image search engines. These are search mechanisms in which the input information determining the user's query is mainly an image, not text. These types of services are embedded in classic search engines as functions or as separate solutions.
Google image search
As we can read on the service page, Google Image Search is a feature called the innovative way of searching in Google's browser. It premiered for the first time in July 2001. Thanks to a search engine specially prepared for this purpose, you can look for information on objects visible in the picture or image.
The image search service is designed to enable all information seekers to find the issues that interest them based on the image and not just a phrase or keyword. For this reason, Google has decided to offer four separate ways to search for information by image or photo.
The first is the "drag and drop" option. It gives you the ability to use an image or photo found directly on the internet. By dragging the image to the search field, you can add it to the search engine and thus receive a list of Google search results.
The second way is to send a photo or image directly from your computer. By simply clicking the camera icon (located in the search box) and selecting "Send image".
The third option is associated with the URL of the image. Based on the web address, of the graphic we can search for more information about it. Here, Google requires users to right-click on the image to copy the image URL and move it to the search box. There, after pressing the camera icon, select the "Paste image URL" option.
The last way to search is to use the extension available for Chrome and Firefox. After installing it, all you have to do is right click on the searched image and Google will automatically search for information about it.
Image search and copyrights
All graphics placed on Google are protected by copyrights. This means that the author's consent is required for use on their own websites. However, it turns out that there is an option for people who need photography for commercial purposes.
Advanced image search has options related to the search for images and graphics with the possibility of using them on other websites. This is guaranteed by the last "Usage Rights" option. When we expand the slider there, we get five options related to the license to use graphics available on Google. In addition, the Google image search option can search for images that are freely modifiable, both for private and professional use.
In addition to searching for the license type, advanced image search will also allow you to find images by subject, size, aspect ratio, region and many others.
Google lens is a feature available in the mobile version of the Google Graphics search engine, as well as in the Google application.
Its function can recognize objects and provide information about them. When aiming the Lens at, for example, a historic object - we will get information about it, a description of the history and what hours and days it is possible to visit. This is something that can be especially useful for vacationers during holidays. Google Lens can recognize the text as well. Just mark a specific line in the restaurant menu to find out more about the meal you want to order. You can also point the camera at the text to add events to the calendar, make a call or translate something from a foreign language.
The Google Lens can also search for objects found in photos. How does this work in practice? Let's assume that we are looking for inspiration for the interior of the living room. We found a photo where we liked the coffee table. We can now choose the "Google Lens" button and then select the table we like in the photo. Google will give us search results based only on the selected table. Thanks to this, even without knowing the name or manufacturer of the furniture, we can immediately learn more about it.
By default, the Google Lens automatically selects the element from the photo on which it will focus, but we can change the object of interest. In some cases, Google recognizes objects and suggests what to search for. In other cases, we have to draw a frame around the object that interests us.
The system is not perfect and is not always precise at recognizing the elements in the pictures. However, it is still being improved so it’s just a matter of time.
Since the Google Lens works directly in the browser, it can be run on any mobile device both Android and iOS. There is no need to update it - the function should be visible to every user.
Better understanding of the consumer
Image recognition technology allows better understanding of the consumer - analysis of photos or videos posted by a specific users allows them to determine which products and which brands they use. It also gives us information how much a user spends on them - e.g. whether they are standard, premium or luxury goods. We can learn what the customer is interested in, what he likes to do, what his lifestyle looks like - this information can significantly help to match the right advertising message to a given recipient at a given moment.
For example, by analyzing a photo from a holiday trip of a particular consumer, we can find out what type of shirt, shoes and which brand he wear, which drink or type of alcohol he prefers (algorithms can also analyze the color of the liquid or the type of glass in which it is served), which type of cuisine and what restaurants he likes and what activities he usually undertakes. Thanks to this information we can see who the consumer really is and not who he is only declaring to be. Also, there is possible a much better match including the place and form of the personalized advertising message displayed in real time.
Image Recognition Technology
Image recognition is a term for technologies that can recognize certain people, animals, objects or other targeted subjects through the use of algorithms and machine learning concepts.
The first stage of operation is obtaining the image. Then follows the analysis and definition of image features. This is a very important stage, because the image description algorithms cannot (generally) struggle with the millions of pixels that make up the original image, but they need to get a convenient interpretation of the image representation in the form of a single point in the feature space.
The next step is to find the right description of the image in the form of a properly selected mathematical formula. Finally, the image is properly recognized and the decision associated with it is made.
This is a classic approach to the topic of image recognition classification. Currently, there is a used network model where the selection of essential features is made by itself at the stage of learning this network (the network learns how to select features and their classification).
Image Recognition usage
One of the applications is facial recognition in public spaces. The use of systems is used, for example, in stores, using CCTV cameras to minimize losses associated with petty theft, recognizing the faces of people previously involved in crime. Similar systems are used at airports, where systems are able to detect the faces of people who should not board the airplane for various reasons.
Another application can be Augmented Reality - here we are talking not from two-dimensional graphics or photos but entire 3D models.
In e-Commerce, mobile applications allow the consumer to take pictures of products and automatically purchase them in an online store.
It is worth remembering that these are only some of the industries and sectors in which automatic image recognition technology is being developed. Machines also learn sounds (including speech), diseases (to support medical diagnostics), economic situation (to optimize financial decisions), geological formations (to find new deposits of raw materials), and social moods (to predict election results). Artificial intelligence will soon dominate in almost all areas of life.
A breakthrough in e-commerce
The customer often buys what looks good at the picture, not really knowing the facts about the product. In e-commerce, marketing using product photos and visual descriptions work best. Many e-stores already use the recommendation engine, which tells customers what other users have bought or viewed. However, global e-commerce giants focus on a more precise system based on image recognition technology.
The system uses an algorithm that compares photos of all products in the store with each other and analyzes their degree of similarity. So if a customer is looking at a specific product, like for example black, woolen pants, the recommendation engine will show them all other similar-looking pants from the entire store's range - starting with those of the closest cut and color. Thanks to this, the user will find what he is looking for more easily. The search process becomes simpler and more user-friendly, which increases the chance that customer visits to the store will end with making a purchase.
Ideas for use of Image Recognition in e-commerce
One of the ideas for using Image Recognition may be a quick search feature. Shopping inspirations can be found everywhere. They are more difficult to accomplish by browsing the internet in search of a dress similar to the one our favorite idol was wearing. Image recognition will help to solve this problem. All that is needed to do is to send a photo of the product using, for example the Messengers integrated chatbot. The system will do the rest, filtering the entire range of stores using the service.
The intelligent image recognition system can be successfully expanded to the level of a virtual stylist who will prompt the client to other, complementary products, well matched to what is already in the basket or a virtual interior designer who will suggest which pillows best suit the selected sofa.
Another benefit that is worth mentioning are new sale channels. Many websites publish dozens of photos, and each of them, regardless of whether they present the celebrity, model or hero of the article - any of them can become an inspiration for shopping. Under the photo may appear a recommendation where to buy products similar to those that are shown. For store owners, this is a chance to launch a new sales channel and redirect traffic to the store. For publishers it is the possibility of additional commercialization of space. Recommendations of this type can already be found on several popular portals, but at the moment they are prepared manually. Thanks to image recognition, the whole process can be automated.
There is also a “tap to buy” feature available - image recognition technology can be used not only for static photos, but also for video materials. Similarly to solutions that can already be found for example on Instagram - all you have to do is hover your mouse over the selected area of the movie and the system will tell us where you can buy exactly this product or very similar to the selected one. This is the next step to shorten the shopping path to the maximum, launching new sales channels, using shopping inspirations and context shopping.
It’s worth mentioning the idea of interactive marketing - there are solutions that identify facial expressions of the users and emotions towards a selected item. When marketers present a brand, logo or product the software can read the emotions of the user (for example if he is smiling, irritated, bored or indifferent). This helps adjust marketing campaigns and predict the reactions of potential customers.
Social media commerce is another case where image recognition technology might come in handy. Facebook for example can recognize objects and places - it can differentiate between foods, objects and facial expressions. The algorithm goes beyond recognizing faces in pictures - Facebook recognizes images of products customers likes and suggests ads accordingly.
Image recognition technology can help detecting and removing inappropriate content on e-commerce sites - like finding and eliminating fake items. Brands and online marketplaces have struggled in the past to find an effective solution. Now it can be done through logo recognition method in which the legitimate brand can find fake logos of counterfeit products and remove any inappropriate or explicit content falsely associated with that brand. As soon as a fake is detected, the item is flagged - this is an automated process which takes away the need for manual input.
In the near future, image recognition technology will find an even wider application in marketing and become more and more important not only because it significantly improves marketing processes, but also because online communication is changing to a more graphic form. More and more content on the web is created in the form of photos, graphics and movies than in the form of text. This trend is constantly growing and necessitates the introduction of tools that enable accurate analysis of this type of data.
The potential for using image recognition technology in marketing and especially in e-commerce is huge. By using it in stores, not only will the sales probability increase significantly but also the chance that the customer will make further purchases and will recommend it to friends and family.