PicScout Growth, Author at PicScout

How To Ace Media Monitoring Like A Boss: Oscars Edition

What if the Oscars were determined by the amount of images online of each nominee? Analyzing an actor’s visibility is a precursor to media monitoring their presence — and for this year’s Oscars awards, we did exactly that.

This year’s Oscars is a tight race, with outstanding performances being outplayed by heavy promotion efforts. Trending actors predicted to win have a visual buzz — the number of images often correlates with their fandom, media support, and various marketing strategies by the film’s producers, their agency, and others.

So we asked ourselves, does visual presence affect the eventual winners of the 88th Academy Awards? How would it affect the voters’ perception of each actor, each category? Are the awards based on merit, politics, or marketing? As the voting by members of the Academy of Motion Picture Arts and Sciences closed on Tuesday, we applied some image-based media monitoring and tallied the results for the Most Visual Performer Award.

Best Supporting Actor:

It’s a tight race between Mahershala Ali (Moonlight) and Dev Patel (Lion): both would make Oscar history if they take home an Academy Award next week. Patel would become the first actor of Indian descent to win an acting Oscar or Ali would be the first Muslim to win an acting Oscar.

Ali was considered the favorite after winning a Screen Actors Guild award for his role in Moonlight, but now Anglo-Indian Patel has joined the front of the race following his BAFTA victory for Lion.

Winner of PicScout’s Most Visual Performer Award (Supporting Actor): Mahershala Ali

Best Supporting Actress:

It’s widely assumed that the versatile Viola Davis will take home the Oscar this year, since scooping up nearly every other award in the season for her moving performance in Fences. By media monitoring her presence, her visual buzz leading up to the Academy Awards is close to 40% — nearly double that of the runner-up, Nicole Kidman.

Winner of PicScout’s Most Visual Performer Award (Supporting Actress): Viola Davis, no competition.

Best Actor:

Visually, it’s a battle of the heartthrobs — Ryan Gosling (La La Land) and Andrew Garfield (Hacksaw Ridge) are in close sights for the Most Visual Performer Award. While neither of them are considered to be serious contenders for the Academy Awards, both of them are closely linked to the visual favorite of the awards season — Emma Stone. Gosling was her La La Land co-star and Garfield has been her on-again, off-again partner for over 7 years. With today’s clickbait celebrity obsession, their visual presence while media monitoring isn’t so surprising.

What is surprising is how they compare to the low image results of the two actual contenders for the award — Denzel Washington (Fences) and Casey Affleck (Manchester By The Sea).

Winner of PicScout’s Most Visual Performer Award (Best Actor): Ryan Gosling, thanks to Emma Stone and swooning fans.

Best Actress:

Emma Stone. Emma Stone. Emma Stone. La La Land’s lead actress dominates the visual buzz surrounding the 89th Academy Awards: she takes up 50% of the images found in her category of Best Actress (runner-up Natalie Portman was next in line with 21%). In more comprehensive media monitoring, Stone also featured in a fifth of all images found of every nominee across lead/supporting actor/actress categories in 2017. Our money, as well as those from the La La Land team, is on her.

Winner of PicScout’s Most Visual Performer Award (Best Actress): Emma Stone.

Best Film:

La La Land has been the Oscars front-runner since its release — it tugs on the heartstrings and nostalgia of Academy voters with music, color and Hollywood dreams. Its visual dominance across the web and social media peaked at the start of the awards season and as “front-runner backlash” kicked in, its visual presence decreased as it neared tonight’s Oscars ceremony. Still, La La Land will take both the Academy Award and PicScout’s Most Visual Performing Film award — its the kind of movie that awards bodies absolutely love.

Winner of PicScout’s Most Visual Performing Film: La La Land — could it be anything else?

FUN FACT

There are thousands more images of our Best Actresses and Best Supporting Actresses than their male counterparts.

Images courtesy:

Viola Davis, Emma Stone, Andrew Garfield © Gage Skidmore

Mahershala Ali, Dev Patel © Gordon Correll

Nicole Kidman © Eva Rinaldi

Ryan Gosling © Georges Biard

Natalie Portman © Josh Jensen

Can Image Recognition Technology Help NBA Stars?

Celebrating NBA’s All-Star Weekend, we’ve put our Visual Insights to the test and held our own MVP awards for the Most Visual Player online. Using our image recognition technology, include our facial recognition capabilities, we compared images found across both traditional and social media and have showcased our top three winners below.

These men play a high-level game with millions of dollars-worth of responsibilities to the NBA and their sponsoring brands. A number of factors influence a brand-ambassador’s visual presence, and however carefully curated it may be, the player’s skills and organic support from fans has a decisive influence.

In applying image recognition technology to measure their visual presence, NBA teams, sponsors and the players themselves can be privy to a range a range of opportunities including:

- brand metrics
- untapped market sectors
- influencer marketing
- evalutating the ROI of these brand ambassadors

Images courtesy:

LeBron James © Martin Ellis
Kevin Durant © Leonard Zhukovsky
Steph Curry © natursports

Globes.com Interview with CEO Uri Lavi

Below is a translated copy of PicScout’s CEO Uri Lavi discussing Visual DNA, visual content analysis and artificial intelligence with Israel’s leading business publication, Globes.

Hebrew link here.

To learn more about Uri’s vision, check out his Medium profile for the latest articles, and contact us to discuss your company’s Visual DNA.

Does Face Recognition Technology Have A Commercial Future? [infographic]

Facebook tagging, Snapchat filters, airport scans: face recognition is increasingly becoming a large part of our everyday lives. Explore below how this innovative technology is helping companies expand their customer-base with AI-driven innovation, giving them the competitive edge in today’s visual world.

Industry Uses Of Face Recognition

Security & law Enforcement

Facial recognition allows security companies and law enforcement agencies to monitor suspicious or undesirable persons. Faces are compared against large databases, and authorities are alerted when there’s a match.

Airports

Airports across the world are exploring the ways facial recognition technology can be used as a secure measure of authenticity to be used instead of passports.

Brands & PR Agencies

Facial recognition helps brands and agencies manage brand ambassadors and high-value assets.

Financial Authentication

Credit card companies are introducing smartphone selfies to authenticate paying for bills. It’s also speculated that mobiles may include selfie snaps as another mode of authentication to unlock phones.

Targeted Advertising

Advertising agencies, for example, have recently been exploring how face recognition technology can assist in targeting their commercials and helping brands achieve their competitive edge. Today, at gas stations and in malls, screens scan customers’ faces in order to run tailored ads with demographic-specific content. Another creative commercial use is fast-food chains offering personalized menus based on facial recognition.

Learn More About This Disruptive Technology

These are just a handful of examples where face recognition is being utilized more and more. On a smaller level, facial recognition technology can help marketers and photo agencies monitor their assets and campaigns – either across the web and social media, or even internally across large photo collections.

Face recognition technology is just one of the many capabilities of PicScout’s Visual API. The way we do business in today’s visual world is changing. Don’t get left behind – talk to us today.

My Twin Celeb, Facial Recognition Tech, Oh My!

Facial Recognition Fun

What a week it was. We launched our new facial recognition app, My Twin Celeb, in preparation for Doppelganger Week 2017 (January 29 – Feb 3) and it quickly became popular through a variety of channels.

My Twin Celeb is a mobile app that features PicScout’s facial recognition software. Within seconds, a user matches a picture of themselves with that of their ‘twin’ celebrity, ready to be shared with friends.

It quickly became viral with thousands of downloads each day, featured on a number of tv programs and even national newspapers getting in on the fun! Making facial recognition technology fun and accessible is one of the best approaches to educate the public — yet we wanted to take this viral app even further.

We thought, how could we measure the visual reach and engagement surrounding the buzz? And what generated it? So we put PicScout’s Insights for Business to use and gathered all the visual data surrounding the hype — both from the web and social media — data that is often left off the marketer’s dashboard. The visual insights we received helped us make better informed decisions in reaching new audiences for the app, offered more relevant images to market it further and gave us new ideas for upcoming apps that showcase our Visual Discovery technology.

Ask us today how our Insights for Business can help market your company’s visual content and presence. And for more information regarding our Visual Discovery API, contact us here.

How Visual Search Can Help Your Business

People upload and share billions of photos every day: the smartphone has become the everyman’s camera and the cloud has expanded our storage capabilities. Visual content is king. In riding this shift in priority content, Visual Search technology has come into new prominence, dominating media coverage and shaping the digital marketing and eCommerce spheres.

What is Visual Search?

Visual Search is the ability to search using an image, against an inventory of hundreds, if not thousands, of images — and finding the exact match or similar images, without the need for text or data, in the blink of an eye.

In today’s image-driven world, it’s getting more and more difficult to find specific products using a text query. Say you spot something you really love, but you don’t know how to find it or what it’s even called: this image recognition technology lets you find all those things you don’t have the words to describe. Think of it as a Visual Search Engine that allows you to easily access, organize, or recommend your images or products by visual similarity.

Unlike image search (which returns images for a text-based query) or reverse image search (which often relies on metadata to match results), Visual Search derives its results from the supplied image itself.

Visual Search and eCommerce: A Natural Partnership

Retail is a natural application of this technology, offering the utmost convenience and intuitive search experience. Keywords can only take you so far when descriptors like “black dress shoe” apply to hundreds, if not thousands, of products. Shoppers, for example, can snap a photo of a piece of clothing on their smartphones, which is then matched with similar styles available to buy on the retailer’s website.

But this isn’t only limited to fashion. Take lifestyle as another example. Imagine shoppers browsing through images of various interior design themes, and the retailer’s image recognition technology will suggest similar items — homewares, soft furnishings, furniture — based on user preferences and selections. Then when shoppers hover over a particular image, the retailer can scan its product listings for all inventory that match (or nearly match) that image (or portion of the image).

Pairing images with its similar products is just one of the many ways Visual Search is transforming eCommerce. This technology can also help retailers with stock management, converting dead ends into sales by displaying visually similar alternatives.

Clean Your Data

Maintenance work on image banks loaded with thousands of images takes a lot of time and manpower — however Visual Search can cut that down in an instant. With the immediate results of Visual Search capabilities, duplicates and unwanted images can be cleaned regularly from your database.

The Many Applications of Visual Search

Visual Search stops the keyword guessing game to find that drawing, art piece, map, or blueprint in an instant. With the snap of a camera, manufacturers can use the technology as a tool to identify the components of their inventory. Publishers, meanwhile, can use Visual Search to source quality visual content from their photo libraries. And Digital Asset Management (DAM) software can incorporate Visual Search capabilities to organize and curate their customers’ content — visually.

For most businesses, this artificial intelligent-driven technology is the logical next step in a world dominated by visuals and imagery. PicScout, a pioneer in image recognition since 2002, has been using and refining its Visual Search technology since its early beginnings. As part of its Visual API capabilities, PicScout’s Visual Search is unrivaled in its pedigree: over the years, we’ve accumulated a registry of over 300-million unique owner-identified images (the largest of its kind in the world) — giving us the advantage of quality subject matter, better auto-tagging and a huge variety of image choices to constantly improve our technology. Our deep learning is based on high-quality photography, producing more focused and more accurate results than many of our competitors.

For a robust, scalable and fast Visual Search for your business, contact us today to discuss how PicScout’s Visual API can help your business, and change the way you and your customers search.

How Face Recognition Increases Customer Engagement & Satisfaction

A Face in the Crowd: The Basics of Face Recognition Technology

Unless you have an identical twin, your face is one of your most unique characteristics. And its distinguishable landmarks – the exact space between eyes, the precise curve of the cheek, the fullness of the lips – can be analyzed by face recognition technology and converted into an individualized ‘faceprint’: a unique identifying tag much like a fingerprint.

As humans, picking faces out of a crowd is something we’re hardwired to do, but training computers to act in the same way is much more difficult. It takes us a split second to process factors like different facial expressions, the wearing of scarves, hats, sunglasses and makeup, as well as a degree of similarity in looks, age, gender and ethnicity. And so in aligning science fiction with reality, deep learning and artificial intelligence has enabled computers to visually search, find and identify specific faces within large image collections.

How Can This Artificial Intelligence-driven Tech Affect My Business?

The number of businesses incorporating this strategic technology is extensive, and though face recognition software is still used largely for security, other applications are becoming more and more popular, particularly in the retail and hospitality industry.

Let’s look a little closer at some of the ways face recognition technology can impact your business — not only can it increase customer engagement, but many businesses who offer it either as a gimmick or as a serous service, are starting to gain a lot of press. If you’re willing to be creative, face recognition technology can be incredibly successful for your company.

Using Face Recognition in the Security Industry

Face recognition technology is now standard at most airports, and aside from law enforcement agencies using the tech to catch criminals, retail is also jumping on the security bandwagon. Wal-Mart, Saks Fifth Avenue and Nordstrom use facial recognition technology to scan the face of anyone who enters a store, identifying any suspicious people and potential shoplifters from their image database, while instantly alerting store security on their phones.

Creative Tech Solutions for Hospitality – Yes, It Works Well!

On some Disney cruise ships, photographers roam the ships for the duration of the vacation taking photos of travelers. The photos are sorted using facial recognition software that pairs the images to the specific people in the photographs. Later, passengers can scan their Disney ID (which they receive before the trip) at a kiosk and all of their pictures are pulled up — without any time being spent searching for yourself in the images.

Face Recognition API Helps Keep Track of Brand Ambassadors

PR firms and photo agencies use facial recognition technology to monitor brand ambassadors and high-value assets, producing results in seconds. Facial recognition and detection helps agencies track campaigns, and the results allow for in-depth analysis with PicScout’s Insights for Business.

The Next Step: Endless Possibilities with Face Analytics

The estimates are that by 2020, the face recognition market will reach $6 billion worth. It’s essential for businesses to stay ahead of the curve with best-practice technology – and in this way, PicScout, a pioneer of image recognition since 2002, offers a broad range of robust, scalable and fast visual capabilities with its Visual API.

Not only does this Visual API use face recognition, but it also applies face detection and face analytics to determine demographics within your visual content. This can be extremely helpful with both content stored on a database, as well as real-time advertising and tracking.

To know more about PicScout’s Visual API and learn how you can apply it to your business, contact us here.

Keeping track of all your images has never been easier, and picking a face in the crowd?

Well, just ask PicScout.

Google Analytics

Google Analytics allows us to measure our websites traffic by viewing statistic reports and getting detailed information about our site’s performance.

Here are a few of the many questions that we can answer about our site by using Google Analytics:

• How many visit the site?

• Where do the visitors come from?

• What directed them to our site?

• Which pages are popular and viewed most?

How to:

• Install Google Analytics by creating an account and signing up.

• Paste the Tracking code you get into the bottom of your HTML content of each page you are planning to track. (Any site that hasn’t been configured yet will say Tracking Unknown until you add the code to your website). This will allow Analytics and your website to talk to one another and interpret information about visits to your site.

• Set up Goals. A Goal is a webpage which a visitor reaches once they have completed an action that you desire. Goals help you make smarter decisions about your design by telling you: Which page was visited most, the geographic location of converted visitors, the keywords that standout in your page and etc.

• For websites with a search box, set up Site Searches. This will track searches made on your website so you can learn more about what your visitors are looking for on specific pages.

• Finally, you will be able to view data by going into Audience Overview Report:

Audience– Gives info about the visitors such as Age, Gender, Interests, Location, Behavior (How often they visit the site).

Acquisition reports– Info about what drove visitors to your website. You will see your traffic broken down by main categories (All Traffic > Channels) and specific sources (All Traffic > Source/Medium).

Behavior reports– Info about your content. Particularly, the top entry pages on your website (Site Content > Landing Pages), and the top exit pages on your website (Site Content > Exit Pages).

Google Analytics

You can also learn how fast your website loads (Site Speed) as well as find specific suggestions from Google on how to make your website faster (Site Speed > Speed Suggestions).

In Picscout:

By using Google Analytics we can conclude how to improve our site’s content and design, understand where visitors may be losing interest and falling off the path along the way (known as “Pain Points”), and learn how to convert more visitors.

Deploying Entity Framework Model First

While working on a project that uses Entity Framework, we noticed that the auto-generated script, created from the Model, overwrites the old scheme and moreover, erases all the data.

Since the project was relatively small and the team working on it was not big, this was not a problem. However, once the project started to be manually tested by QA, we had to find a simple deployment process to update the DB without restarting it.

The problem:

We didn’t want to use EF migrations because of two reasons:

It requires to do each change twice: First – in the DB, and Second – in the code.
We didn’t want to manage every small change.

The solution:

We found a simple and easy solution that satisfies our needs:

Step 1

We ran the auto-generated script on a side DB (that will be erased after the process)

So at that point we had 2 DBs:

Old DB – the original DB which exists in production and isn’t updated.

New DB– the new and temporary DB which has the updated scheme.

Step 2

We executed the SqlPackage command line tool (part of SQL Server) on the New DB and created a .dacpac file.

A “dacpac” is a file with a .dacpac extension, which holds a collection of object definitions that one could find in a SQL Server database such as tables, stored procedures, views etc.

Step 3

Using the created .dacpac file we executed the Publish command on the SqlPackage tool, which compares between the two DBs and allows actions such as add/remove/update on fields, types, SPs and more.

The main problem with the publish command is with the implementation that could cause a serious performance issue. In some cases, the original table is copied to a side table with all the data, then a new table is created with the new scheme and lastly, the data from the side table is copied into the new table. Therefore, in case of a large data set it could take a long time.

Summary

DB migrations is a known issue, and there are a lot of good solutions out there. In our special case we decided to use a simple and easy solution which can be implemented with basic tools that arrive with the SQL Server version we had. This solution does not fit in any situation, however, for us it did the work.

Writing Web-Based Client-Side Unit-Tests with Jasmine and Blanket

Preface

When writing a website, or more often – a one page app – there is a need to test it, just like any other piece of code.

There are several types of tests, of course, including unit tests and integration tests.

While integration tests test flows of the entire application, end to end, and thus simulate user interaction (which requires special browser-based testing package), unit tests run specific functions.

However, when writing an entire application in JavaScript, running pieces of code is a bit more tricky.

On one hand, we are not used to writing unit tests in JavaScript, and run the tests completely in the browser. On the other hand, calling JavaScript code and then testing various members for values is much more easily done when written directly in JavaScript.

Luckily, the good people of the web has given us several JavaScript based packages for writing unit tests. I’ll talk about Jasmine, and add some words about Blanket, that integrates with Jasmine to perform code-coverage.

Jasmine

Jasmine is a JavaScript based library to perform unit tests. It consists of several parts:

The Runner
Tests Framework
Plug-ins

1. The Runner

The runner is an HTML file with base code that loads the tests framework and runs the tests. It will not do anything when you take it out-of-the-box. You have to add your own scripts in there, so consider it a template.

The base HTML looks like this:

<link rel="shortcut icon" type="image/png" href="jasmine/lib/jasmine-2.0.0/jasmine_favicon.png">
<link rel="stylesheet" type="text/css" href="jasmine/lib/jasmine-2.0.0/jasmine.css">

<script type="text/javascript" src="jasmine/lib/jasmine-2.0.0/jasmine.js"></script>
<script type="text/javascript" src="jasmine/lib/jasmine-2.0.0/jasmine-html.js"></script>
<script type="text/javascript" src="jasmine/lib/jasmine-2.0.0/boot.js"></script>

Next, you need to add your own application scripts:

<script type="text/javascript" src="src/myApp.js"></script>

And finally comes your tests scripts:

<script type="text/javascript" src="tests/myAppTestSpec.js"></script>

2. Tests Framework

Jasmine have several files that create the tests framework. The most basic ones are the one listed above, in the basic HTML example. Let’s go over them quickly:

jasmine.js

The most basic requirement. This is the actual framework.

jasmine-html.js

This one is used to generate HTML reports. It is a requirement, even if you don’t want HTML reports.

boot.js

This one was added in version 2.0 of Jasmine, and it performs the entire initialization process.

Writing Tests

Structure

The unit tests in Jasmine are called “Specs”, and are wrapped in “Suites”. It look like this:

describe("A suite", function() {
  it("contains spec with an expectation", function() {
    expect(true).toBe(true);
  });
});

The describe function describes a test suite, while the it function specifies a test.

Note that those two get as parameters a name and a simple function block, and that the it block is being called in the body of the describe function block. This means you can store “global” members for each test suite. Also it means that the tested code comes inside the it block, along with any assertions.

Expectations (a.k.a. Asserts in other test suites)

When writing a unit test you expect something to happen, and you assert if it doesn’t. While in other test suites you usually use the term Assert to perform such operation, in Jasmine you simply Expect something.

The syntax for expectations is straight forward:

expect(true).toBe(true);

There are many “matchers” you can user with the expect function, including but not limited to:

toBe – test the value to actually BE some object (using ‘===‘).
toEqual – test the value to EQUAL some other value.
toMatch – tests a string against a regular expression.
toBeDefined / toBeUndefined – compares the value against ‘undefined‘.
toBeTruthy / toBeFalsy – tests the value for JavaScript’s truthiness or falsiness.
toThrow / toThrowError – if the object is a function, expects it to throw an exception.

You can also negate the expectation by adding not between the expect and the matcher.

Spies

You can also use Jasmine to test if a function has been called. In addition, you can (actually, need) to define what happens when the function is called. The syntax looks like this:

spyOn(someObject, "functionName").and.callThrough();
spyOn(someObject, "functionName").and.returnValue(123);
spyOn(someObject, "functionName").and.callFake( ... alternative function implementation ... );
spyOn(someObject, "functionName").and.throwError("Error message");
spyOn(someObject, "functionName").and.stub();

Then, you can check (expect) if the function was called using:

expect(someObject.functionName).toHaveBeenCalled();

expect(someObject.functionName).toHaveBeenCalledWith(... comma separated list of parameters ...);

More info

There are many features you can use with Jasmine. You can read all about it in the official documentation at http://jasmine.github.io/

3. Plug-ins

Well, I’ll only talk about Blanket, the code coverage utility that integrates with Jasmine.

In the runner, add the following line before tests specs scripts, but after the application scripts:

<script type="text/javascript" src="lib/blanket.min.js" data-cover-adapter="lib/jasmine-blanket-2_0.js"></script>

and that’s it!

Below the test results report there will be the code coverage report.

The blanket.js package can be found at http://blanketjs.org/ and the adepter for Jasmine 2.x can be found at https://gist.github.com/grossadamm/570e032a8b144ec251c1 (unfortunately, blanket.js only comes pre-packaged with an adapter for Jasmine 1.x).

Happy Coding!