Google Lens: Let’s See How it Searches What We See

What is Google Lens?

Google Lens is an image recognition system that uses visual analysis and a neural network to bring up relevant information about the items it recognizes. It identifies an object when a user points their phone’s camera at the object by analysing an image, barcodes, QR codes, labels, and text. It then presents relevant search results, web pages, and information. For instance, the device will immediately join the scanned Wi-Fi network when the camera is aimed at a Wi-Fi label with the network name and password. Google Lens works with Google Photos and Google Assistant as well. The program’s functionality is analogous to Google Goggles, a predecessor that performed a similar function but with fewer features. Google Lens uses more sophisticated deep learning techniques to enhance its detecting abilities than its earlier version.

What is Google Lens Used For?

  • Translate: With Google Translate installed, a user can point the phone at text and have it live-translated in front of his eyes. This may also be done without an internet connection.
  • Smart Text Selection: The user can use Google Lens to highlight text and copy it to use on the phone after pointing the phone’s camera at it. As an example, consider being able to point the phone at a Wi-Fi password and copy/paste it into a Wi-Fi login page.
  • Smart Text Search: When highlighting text in Google Lens, you can use Google to search for that text. This is useful if the user needs to search for a word’s definition, for instance.
  • Shopping: If the user sees a dress while shopping, Google Lens can recognize it and other apparel items identical to it. This works for almost every item a user can think of, as well as for buying and reading reviews.
  • Google Homework Questions: That’s true, the user can simply scan the query to see what Google returns.
  • Searching Around: Google Lens will recognize and identify the user’s surroundings if the user points the camera around him. This might include information on a landmark or information about different sorts of cuisine, including recipes.
  • Copy Text from the Real World: The capacity of Google Lens to capture text from a physical document — a paper, a book, a whiteboard, a suspiciously wordy tattoo on the rumpus, or anything else with writing on it — and then copy that text into your phone’s clipboard is its most powerful and most often used feature. The user can then effortlessly copy and paste the content into a Google Doc, a note, an email, a Slack discussion, or anything.
  • Send Text from the Real World to Your Computer: Just follow the same procedures as before, but this time, scan the panel at the bottom of the screen for the “Copy to computer” option. Any machine running Windows, Mac, Linux, or Chrome OS should have that choice available as long as the user is actively signed into Chrome with the same Google account. And when the user taps it, a list of all the locations that are accessible is prompted on the screen.
  • Hear Text from the Real World Read Aloud: Simply repeat the previous procedure of pointing the phone at the paper and selecting the “Text” option. Choose any text, and this time, seek the little “Listen” option in the panel at the bottom of the screen.
  • Interact with Text from an Image: In addition to the live stuff, Lens can pull and process text from images — including both actual photos and screenshots the user has captured. That latter part opens up some pretty interesting possibilities. Say, for instance, the user has just gotten an email with a tracking number in it, but the tracking number is some funky type of text that annoyingly can’t be copied, or maybe the user is looking at a web page or presentation where the text for some reason isn’t selectable. Well, grab a screenshot — by pressing your phone’s power and volume-down buttons together — then make the way over to the Google Lens app. Look for the screenshot on the Lens home screen, tap it, and tap “Text” along the bottom of the screen. And then, the user can simply select the required text. From there, you can copy the text, send it to a computer, or perform any of Lens’s other boundary-defying tricks.
  • For Design Patent Search: Google lens is a helpful tool to conduct an effective design patent search. Extensive design patent searches are a typical and essential method for assuring that the design application is granted. Although applicants can carry out design searches on their own, it is recommended to seek professional help in order to get accurate search results and avoid the risks of rejection.

The architecture of Google Lens

Google Lens uses cloud Vision API to find products of interest within images and visually search product catalogues.

Object Localizer

The Lens may utilize the Vision API’s “Object Localization” option to identify and gather data about various items in a picture. The following items are returned for each object that is detected:

  • What is a textual description, exactly, in everyday terms?
  • A confidence score indicates how confident the API is in what it has found.
  • The vertices of the enclosing polygon around the item were normalized. Where do the things appear in the picture?

Product search

Retailers may design items with reference photos that visually explain the product from a variety of angles using the Vision API Product Search. Then, retailers may include these items in product sets. The following product categories are supported by Vision API Product Search at the moment: general, packaged goods, toys, clothes, and home goods.

Vision API Product Search uses machine learning to match the product in the user’s query image with the photos in the retailer’s product set when customers query the product set with their own photographs. It then produces a ranked list of visually and semantically related results.

Cloud storage

Cloud storage provides the already stored image information which can be used for object identification, text detection, and translation. The URL of a file located in cloud storage or a base64-encoded image string is the two ways to transmit an image to the Vision API for picture detection.

Algorithmic Solutions

Google launched Google Lens a few years ago to lead the drive for “AI first” goods. Now that machine learning techniques have improved, particularly in the areas of image processing and natural language processing, Google Lens has reached new heights. Here, we examine a few algorithm-based approaches that underpin Google Lens.

The Lens enables users to convert what they see in the real world into a visual search box, allowing them to identify items like plants and animals or copy and paste text from the real world onto their phones.

Region Proposal Network

Google Go’s Lens must interpret the shapes and characters in a picture after it has been captured. Tasks involving text recognition depend on this. In order to find character-level bounding boxes that may be combined into lines for text recognition, optical character recognition (OCR) uses a region proposal network (RPN).

A fully convolutional network RPN predicts object limits and objectness scores at each place at the same time. Fast R-CNN performs detection using high-quality region suggestions that RPN has been trained to provide. It essentially instructs the unified network where to search.

Knowledge Graphs

A picture with a bounding box around text that can be identified is on the left. This image’s initial optical character recognition (OCR) result states, “Cise is beauti640.” The phrases “life is lovely” are recognized by Lens in Google Go by using Knowledge Graph in addition to the context from neighboring words, as seen on the right.

There might be a variety of extra difficulties because the photos that Lens in Google Go captures could come from sources like signs, handwriting, or papers. The model may misread words if the text is veiled, stylized, or fuzzy. Lens in uses the Knowledge Graph to give contextual hints, such as whether a term is probably a proper noun and shouldn’t be spell-checked, in order to increase word accuracy.

Convolutional Neural Networks (CNNs)

Convolution neural networks (CNNs) are the foundation for many computer vision applications because of the availability of massive datasets and computing power. As a result, the focus of the deep learning community has primarily switched to boosting CNN performance in image identification.

To identify cohesive text blocks, such as columns, or text with a recognizable style or colour, Lens uses CNNs. The final reading order of each block is then determined inside each block using signals such as text alignment, language, and the geometric connection between the paragraphs.

Separable convolutional neural networks (CNNs) with an extra quantized long short-term memory (LSTM) network carry out each of these tasks, including script detection, direction identification, and word recognition. Additionally, a range of data, including ReCaptcha and scanned pictures from Google Books, are used to train the models.

Google Lens Alternatives and Similar Apps

1. Search by Image on the Web

Reverse image searching has gained a lot of traction. The user may choose a picture from the device’s gallery or capture a shot with the camera and do reverse image searching using the Browse by Image on the Web application. This aids in locating the photographs that the user wants to look for if they are accessible in high resolution or if the user wants to learn more about the item. If the user wants to search for a certain product inside an image, the user can also utilize the Crop Feature on the Search by Image on the Web app to alter the picture. Be aware that there are several advertisements that might make the experience with reverse image searches difficult. This program lets the user search the Web for things, pictures, or people. The software may assist the user in determining if a social network profile photograph is genuine or fraudulent, locating related images, learning more about a clothing item, and discovering where to purchase it, among other things. Additionally, the user may scan QR Codes with the integrated QR scanner software.

2. Reverse Image Search & Finder

One of the well-known and trustworthy reverse image search programs, Reverse Image Search & Finder enables a user to look for photos from any source. The user may start the reverse image search by taking an image with a camera or selecting any image from the gallery. Using the URL or link for the picture, the user may also conduct a reverse image search. To extract the precise picture, the user may crop, rotate, and more using the built-in photo editor in the Reverse Image Search & Finder tool. The user interface of the app is sleek and contemporary. It is uncluttered and one of the simpler Google Lens competitors because there aren’t many buttons.

3. Image Analysis Toolset (IAT)

The Image Analysis Toolset is one of the top-rated applications on the Play Store that assists the user with reverse image search, item recognition, and much more (IAT). A wide range of categories, including inanimate things, plants, and animals, are supported by the app. The IAT app might be useful if the user wants to learn more about any object or picture. It will display details about the picture and the things in it, including labels, links to relevant web pages, and matching and aesthetically related pictures (if available). The Censorship Risk Meter, which tells the user if a picture could be banned or censored, is built-in.

Additionally, there is an optical text recognition capability. The user may quickly alter and add anything that he wants to the scanned document by using it to extract the text from any document. A logo recognizer, landmark recognizer, barcode scanner, colorimeter, and face insight to determine facial characteristics, emotions, and degree of likeness are a few further features.

4. PictPicks

A user may use the PictPicks app to look for pictures that are similar to the ones the user has submitted. Additionally, the app features a built-in filter that enables the user to block explicit material and do more detailed searches. The user may begin an image search using PictPicks’ Search By Picture tool by selecting an image from the phone’s gallery or one that was taken with the camera. The app offers options to share, store, and utilize photographs in other applications, as well as to display all relevant information about the searched image, including image sources.

The software has a highly cutting-edge and sophisticated user interface. The learning curve for PictPicks is really low, and the user can do a reverse image search straight away. Notably, the program occasionally displays advertisements without interfering with the work. Overall, PictPicks is among the most effective Google Lens substitutes.

5. PictureThis

Notably, a user can use this software to recognize any type of plant, including trees, flowers, succulents, cactus, and more. Additionally, if you have a backyard garden and need advice on how to take care of the plants, this app may be a tremendous asset. The Image With an accuracy rate of 98%, this software can identify more than 10,000 plant species.

If the user wants to gain more in-depth information about plant care, one may also look into a network of helpful horticultural experts. The software offers an intuitive user interface, and instructions to help the user snap the right picture, and, if the location is enabled, it can find nearby plants. In the Feeling Lucky area of the app, there is a mini-game that rewards users with free plant identification bonuses.

Developments with Respect to Social Media

Future-oriented visual search is a major trend that will probably alter how we find and buy pertinent things. In order to make it easier for consumers to locate exactly what they’re looking for, Snapchat has been discussing this for years, Pinterest is expanding its visual search functionality, and Facebook is also investigating image-based solutions. The creation of Google’s visual search capabilities is crucial since no platform has invested more in search than Google.

The Broadened Scope of Google Lens

The scope of Lens’s potential has increased. In contrast to objects and portraits, Google estimates that up to 15% of photographs taken by people nowadays are of shopping lists and receipts. Lens now includes the capability to recognise text in photographs and convert it for use in other apps in order to address this. With an increasing number of applications and a foundation built on Google’s vast database, the Lens offers a lot of potential for development. Visual search is another crucial consideration to make and will become much more crucial in the future of SEO, even if the rise of voice search, particularly as a result of the development of smart home devices, has garnered the majority of attention.

Conclusion and Future Scope

Google Lens with AR glasses will change the current scenario of seeing the things around the users. Currently, the user can use a phone to get information about a subject by pointing the camera at it. Google is driven to create a system that combines Google AR glasses with Google Lens. By eliminating intermediary obstacles like phone cameras, this technology will enable users to instantly get the necessary information. For instance, a person who only speaks English might find themselves in a Chinese restaurant with a menu in that language. While the user is looking at the menu, the Google AR glasses will instantly translate each item from Chinese into English. To bolster this, Google published a blog post describing the development of its Lens picture search feature and outlining the various improvements and developments they made to make it a more practical, precise tool.

References

This Article was originally published here :- https://www.copperpodip.com/post/google-lens-is-searching-what-you-see

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Copperpod IP

Copperpod IP

Copperpod is one of world's leading intellectual property research and technology consulting firms.