{"id":41241,"date":"2020-07-21T15:14:29","date_gmt":"2020-07-21T13:14:29","guid":{"rendered":"https:\/\/www.sotrender.com\/blog\/?p=41241\/"},"modified":"2021-06-18T08:59:59","modified_gmt":"2021-06-18T06:59:59","slug":"ai-emotion-recognition","status":"publish","type":"post","link":"https:\/\/www.sotrender.com\/blog\/2020\/07\/ai-emotion-recognition\/","title":{"rendered":"Using artificial intelligence and emotion recognition for PR"},"content":{"rendered":"

Businesses recognize that several factors play an important role in eliciting positive emotions about their brand. Who represents the brand and how can make or break your relationship with your customer base. Those that come across as more human and real should give off a more positive vibe to their customers. Given the chance, brands would benefit from analyzing how their representatives come across emotionally.<\/span><\/p>\n

How can you do that? By using artificial intelligence, of course! We put our emotion recognition model to the test by analyzing how two entrepreneurs express their emotions on their Instagram profiles to their millions of followers.\u00a0<\/span><\/p>\n

<\/p>\n

Emotion recognition software and models<\/strong><\/h2>\n

Aside from declaring your emotions verbally, there are other ways that we can let each other know how we\u2019re feeling. Your muscles might spasm, you might start looking away, or start slouching. These are visual cues that give us a hint that you might be feeling a certain way.<\/span><\/p>\n

AI has a long way to go before it can perfectly recognize emotions<\/b>. People are still debating Ekman\u2019s classic theory of emotions – which were used as a basis for future artificial intelligence and machine learning projects. The theory, in essence, states that there are six universal human emotions and that you can find them expressed similarly in different cultures. Critics argue that the current AI technology can only process exaggerated expressions and that their training sets are not diverse enough.<\/span><\/p>\n

Despite these arguments, there is still merit in using AI to detect facial expressions. We\u2019ll explain how you can use machine learning to extract emotions from images below.<\/span><\/p>\n

How does emotion recognition work?<\/strong><\/h2>\n

The AI will try to determine your emotional expression based on several factors such as the location of your eyebrows, eyes, and how your mouth is positioned. These models detect 8 emotions: neutral, happy, sad, surprise, fear, disgust, anger, and contempt in faces. In case you\u2019re wondering, this is how some of those are classified:<\/span><\/p>\n

\"emotion<\/a> \"emotion<\/a> \"emotion<\/a> \"emotion<\/a><\/p>\n

But thanks to our R&D team at Sotrender<\/a>, we can break down how the emotion extractor works. The emotion extractor system includes two machine learning models: a face detector and emotion classifier model. Don\u2019t worry, we\u2019ve prepared a few examples to help illustrate this point.<\/span><\/p>\n

How can Sotrender\u2019s artificial intelligence models detect faces and emotions?<\/strong><\/h2>\n

Most classical vision algorithms allow machine learning engineers to detect one face in an image. However, some solutions integrated several classical approaches to be able to recogni<\/span>ze several faces and their emotions in a single image. In our case, we\u2019re using machine learning approaches, which are data-driven, for this task. This is because our model uses state-of-the-art machine learning algorithms, namely <\/span>MTCNN architecture<\/b> to detect faces and <\/span>ResNet50 architecture<\/b> to classify emotions. <\/span>\u00a0How does Sotrender compare to the most advanced model?<\/span><\/p>\n

When tested, the state-of-the-art accuracy on the AffectNet dataset is currently 59.58%, while our model\u2019s accuracy scores at 57.1%. Upon first glance, neither of these screams \u201cexceptional\u201d to those that aren\u2019t familiar with machine learning. Essentially, our model is comparable in accuracy to the state-of-the-art models that are currently available.<\/span><\/p>\n

96% of the state-of-the-art<\/strong>
\n\"\"<\/a><\/p>\n

Visualizing the emotion recognition model<\/h2>\n

The process that these images undergo is actually fairly simple to understand when you visualize it. First off, we need an image with faces. This is the <\/span>input image<\/b> that we\u2019re going to analyze.<\/span><\/p>\n

\"\"<\/a><\/p>\n

The next step is to feed the input image into the face detector model.<\/span><\/p>\n

\"\"<\/a>So far so good! Now, the faces will be <\/span>cropped out<\/b> of the original image to become their own stand-alone images.<\/span><\/p>\n

\"\"<\/a><\/p>\n

And finally we\u2019re at the <\/span>emotion classification step<\/b>. We\u2019ll feed these images into the second model to try to recognize their emotions. The images will no longer be cropped, but instead be presented as the original image with a prediction.<\/span><\/p>\n

\"emotion<\/a><\/p>\n

Sotrender\u2019s model can detect and recognize neutral, happy, sad, surprise, fear, disgust, anger, and contempt as emotions <\/span>in images and videos<\/b>.\u00a0\u00a0<\/span><\/p>\n

To sum it up, the process looks like this:<\/span><\/p>\n

\"face<\/a><\/p>\n

What can we do with this method and what exactly does it tell you about your own company?\u00a0<\/span><\/p>\n

    \n
  • We can quickly show you what emotions are present in your brand\u2019s communication, whether it\u2019s on social media, images, or videos.\u00a0<\/span><\/li>\n
  • You would expect that most of your ads would include people who are glowing, flawless, and happy, so this part of the analysis would be extremely straightforward.<\/span><\/li>\n
  • Finally, you will find out how the \u201cfaces of your brand\u201d are presenting themselves, and in turn, your brand. Whether these are influencers or the CEO of the company acting as <\/span>brand ambassadors<\/span><\/a>, their emotions will definitely influence your target audience<\/a>.\u00a0<\/span><\/li>\n<\/ul>\n

    We can always make adjustments to the models so they can fit social media data. Specific business propositions make it easier to create a dataset that our clients can benefit from. So don\u2019t be a stranger and just drop us a line if you would like us to prepare something that fits your needs.<\/span><\/p>\n

    Emotional intelligence, leadership, entrepreneurship<\/strong><\/h2>\n

    Businesses are engaging in leadership discourse, and this discussion includes topics such as what qualities make a good leader, who should lead, and how. One topic that tends to reappear as a crucial trait in an effective leader is <\/span>emotional intelligence<\/b>. This is because emotional intelligence and being aware of what emotions you are communicating to others can improve your relations with your employees and regular customers.<\/span><\/p>\n

    Just as IQ is traditionally measured with ability tests, the same goes for emotional intelligence. Emotional intelligence is a set of abilities related to recognizing your own and others\u2019 emotions, understanding the differences between these emotions, and knowing how to alter emotional states. There is some <\/span>empirical support<\/span><\/a> that indicates that <\/span>emotional intelligence is related to positive work attitudes, altruism, and work outcomes<\/b>. What\u2019s more, <\/span>other research<\/span><\/a> has shown that the <\/span>best-performing managers were rated as being more \u201cemotionally competent\u201d<\/b>.\u00a0<\/span><\/p>\n

    Businesses and customer bases are made up of people who have their own needs and feelings, so it makes sense that they have expectations about how others should be treated. It seems that the better the employees feel at a company, the better that company will be perceived. Otherwise, people will be incentivized to avoid the brand altogether.<\/span><\/p>\n

    We know that who represents your brand is sometimes almost important as the products themselves<\/b>. Steve Jobs\u2019 following would have told you the same thing. That\u2019s why it\u2019s crucial for the leaders to not just be involved in their business, but also be authentic with their audience. <\/span>Still, most of them understand that they\u2019re subject to scrutiny as much as anyone else if they misbehave publicly (whether intentionally or unintentionally).<\/b><\/p>\n

    Let\u2019s take a look at how two well-known entrepreneurs present themselves and how their behavior is seen by others. In this article, you\u2019ll see how machine learning can provide a better understanding of how and how often we display some emotional reactions.\u00a0 But first, we\u2019ll give you a short intro to both of them.<\/span><\/p>\n

    Richard Branson (Virgin Group)<\/strong><\/h2>\n

    Richard Branson has been active in many of his companies\u2019 promotional campaigns and PR. Most of his acts and performances were seen positively, such as trying to publicize his wedding and bridalwear store by putting on a <\/span>dress <\/span><\/a>and makeup. Though not all of his antics would necessarily come across as great in 2020 – specifically, dressing up as a Zulu warrior.<\/span><\/p>\n