What It Takes to Train The Next Generation of Innovators

What It Takes to Train The Next Generation of Innovators

Hardik Gandhi   January 25, 2018

Unsplash

This article was published on GrowthX Academy’s Blog on August 28, 2017.

Sean Sheppard, founder of GrowthX Academy, discusses the critical skills for the upcoming “Innovation Economy”.

“How do we educate people for a future we can’t predict?” It’s a question that’s been on my mind a lot lately — and, it turns out, it’s been on Sean Sheppard‘s as well.

Sean is a serial entrepreneur, venture capitalist, and the founder of GrowthX and the GrowthX Academy. He’s someone who’s been steeped in modern sales, marketing and growth hacking methods, so I was excited to get the chance to chat with him recently about the skills he believes will be critical for the coming “Innovation Economy.”

The Problem: Our Outdated Education System

Most of us have the sense that our education systems haven’t kept pace with innovation. In our conversation, Sean explains how deeply behind we’ve fallen:

“The modern education system was developed in the Age of Enlightenment to support the Industrial Revolution of the 19th century as a way to take people off of farms and educate them to work in factories. That’s why there are school bells. They’re meant to mimic factory whistles. That’s why we have the people lined up in desks, in rows, because that’s how an assembly line is constructed.”

As Sean notes, this transition was critical. “In the 1900s, 40 percent of the jobs in this country were farming jobs. Today, only 2 percent are farming jobs.” Moving from an agriculture-driven society to an industrial one required education systems that prepared students for the kinds of jobs that were becoming available.

Sean points out that we’re in a similar transition now. “Very soon, 40 to 50 percent of the jobs are going to be replaced by robots and automation. We’re now entering what the World Economic Forum has called the fourth industrial revolution: the ‘Innovation Economy.’”

The Four Factors of Future Effectiveness

So what changes do we need to make to prepare for this coming transition? What skills do students and professionals need to practice today to build competency for future jobs? Sean highlights four pillars in particular that form the basis of his approach at GrowthX (I’ll give you a hint — none of them involve getting an MBA or liberal arts degree).

1. Mindset

I was happy to hear Sean touch on mindset as one of his four pillars, as it’s something I’ve been hammering into my team at Web Profits. Sean and I agree — the future belongs to those who adopt a growth mindset, rather than a fixed mindset.

None of us can predict with 100% certainty what the future of the Innovation Economy looks like (except maybe Mark Zuckerberg). Limiting yourself with a fixed mindset — one that restricts you to considering things as they are, not as they might be — could prevent you from identifying and taking advantage of opportunities as they arise.

That’s somewhat obvious, but Sean added an important note: “There is no distinction between personal and professional development in the Innovation Economy.” You can’t think of your future performance in terms of your career alone. Embracing the growth mindset Sean suggests means recognizing that every part of yourself — from your work to your health and beyond — can, and should, be improved upon.

2. Mastery

Having a growth-based mindset provides needed flexibility for an unclear future. But mindset alone doesn’t fully answer the question of how you prepare today for jobs that may not exist until tomorrow.

That’s where competency-based education comes into play, according to Sean. “Competency-based education models will be the future of education. It’s the idea that we can measure people the same way you and I measure marketing efforts in real-time. We can assess people quickly about whether or not they’ve achieved the competency.”

Out of competency, Sean suggests, mastery grows. “You acquire the knowledge; there’s a framework for that. You practice it to demonstrate that you can acquire the competencies, and then through the repetitive iteration of that, you develop proficiency and then, ultimately, mastery.”

Sean’s model makes more sense when applied to a hypothetical job. Suppose you want to become a growth hacker. There’s no “official” training program; no university you can attend. So how do you prepare for this job? According to Sean, you study the existing knowledge that’s available. You identify and develop the core competencies involved in the job. Then, through practice, iterative improvement and the simple investment of time, you eventually achieve mastery.

The beauty of this approach is that it’s available to everyone. Sean states, “It’s about being a learn-it-all not a know-it-all. It’s about understanding that the foundation of mastery is that you do not have to be born with some natural level of inborn talent or set of skills.”

3. Career

Transforming personal and professional mastery into a career will look different than it used to, according to Sean. “As an individual you have to focus on your career development, and as a manager and a leader, you have to focus on helping people develop their careers.”

Long tenures with a single company are practically nonexistent these days, and our transition to the Innovation Economy will only accelerate this change. Succeeding in this future — in whatever role you define yourself — will require that you take an active role in managing your career, as well as helping guide the careers of others.

4. Community

Mindset, mastery and career are all factors you develop on your own. But, in Sean’s opinion, where things really come together is in a focus on community. “The modern education requires diversity of thought, opinion, background, and experience from a whole host of different points of view.”

Simply put: you need a diverse community whose wisdom you can draw on to advance your learning beyond what you’re capable of on your own.

Sean attempts to build communities like these through GrowthX (the next session starts September 12th), but you can also cultivate your own community by connecting with older mentors, those in other industries and thought leaders you admire.

Now isn’t the time to remain idle. By focusing on updating your mindset, mastery, career and community, you’ll be ready to face whatever challenges come your way in the new Innovation Economy.

---

This article was published on GrowthX Academy’s Blog on August 28, 2017.

---

How to Design Social Systems (Without Causing Depression and War)

How to Design Social Systems (Without Causing Depression and War)

Hardik Gandhi   January 20, 2018

How to Design Social Systems (Without Causing Depression and War)

Here I’ll present a way to think about social systems, meaningful interactions, and human values that brings these often-hazy concepts into focus. It’s also, in a sense, an essay on human nature. It’s organized in three sections:

• Reflection and Experimentation. How do people decide which values to bring to a situation?
• Practice Spaces. Can we look at social systems and see which values they support and which they undermine?
• Sharing Wisdom. What are the meaningful conversations that we, as a culture, are starved for?

I’ll introduce these concepts and their implications for design. I will show how, applied to social media, they address issues like election manipulation, fake news, internet addiction, teen depression & suicide, and various threats to children. At the end of the post, I’ll discuss the challenges of doing this type of design at Facebook and in other technology teams.

Reflection and Experimentation

As I tried to make clear in my letter, meaningful interactions and time well spent are a matter of values. For each person, certain kinds of acts are meaningful, and certain ways of relating. Unless the software supports those acts and ways of relating, there will be a loss of meaning.

In the section below about practice spaces, I’ll cover how to design software that’s supportive in this way. But first, let’s talk about how people pick their values in the first place.

We often don’t know how we want to act, or relate, in a particular situation. Not immediately, at least.

When we approach an event (a conversation, a meeting, a morning, a task), there’s a process — mostly unconscious — by which we decide how we want to be.

Interrupting this can lead to doing things we regret. As we’ll see, it can lead to internet addiction, to bullying and trolling, and to the problems teens are having online.

So, we need to sort out the values with which we want to approach a situation. This is a process. I believe it’s the same process, whether you’re deciding something small — like how openly you will approach a particular conversation — or something big.

Let’s start with something big: many teenagers are engaged in sorting out their identities: they take ideas about how they ought to act (manly, feminine, polite, etc) and make up their own minds about whether to approach situations with these values in mind.

Worksheets from “On My Own Terms”. Join our community to play these games!

For these teens, settling on the right values takes a mix of experimentation and reflection. They need to try out different ways of being manly, feminine, intelligent, or kind in different situations and see how they work. They also need to reflect on who they want to be and how they want to live.

These two ingredients — experimentation and reflection — are required to sort out our values. Even the small decisions (for example, deciding how to balance honesty and tact in a conversation) require experimenting in real situations, and reflecting on what matters most.

This process can be intuitive, nonverbal, and unconscious, but it is vital.¹ If we don’t find the right values, it’s hard to feel good about what we do. The following circumstances interfere with experimentation and reflection:

• High stakes. When deviation from norms becomes disastrous in some way — for instance, with very high reputational stakes — people are afraid to experiment. People need space to make mistakes and systems and social scenes with high consequences interfere with this.
• Low agency. To put values to the test, a person needs discretion over the manner of their work: they need to experiment with moral values, aesthetic values, and other guiding ideas. Some environments — many of them corporate — make no room for being guided by one’s own moral or aesthetic ideas.
• Disconnection. One way we judge the values we’re experimenting with is via exposure to their consequences. We all need to know how others feel when we treat them one way or another, to help us decide how we want to treat them. Similarly, an architect needs to know what it’s like to live in the buildings she designs. When the consequences of our actions are hidden, we can’t sort out what’s important.²
• Distraction and overwork. We also lose the capacity to sort out our values when reflection becomes impossible. This is the major cost of noisy environments, infinite entertainment, push notifications, and some types of poverty.
• Lack of faith in reflection. Finally, people can come to consider reflection to be useless — or to be avoided — even though it is so natural. The emotions which trigger reflection, including doubt and confusion, can be brushed away as distractions. One way this happens, is if people view their choices through a behaviorist lens: as determined by habits, reinforcement learning, or permanent drives.³ This makes it seem like people don’t have values at all, only habits, tastes, and goals. Experimentation and reflection seem useless.

Software-based social spaces can be disastrous for experimentation and reflection.

One reason that private group messaging (like WhatsApp and Messenger) is replacing virality-based forums (like Twitter, News Feed, and increasingly, Stories) is that the latter are horrible for experimenting with who we are. The stakes are too high. They seem especially bad for women, for teens, and for celebrities—which may partly explain the rise in teen suicide—but they're bad for all of us.

A related problem is online bullying, trolling, and political outrage. Many bullies and trolls would embrace other values if they had a chance to reflect and were better exposed to consequences. In-person spaces are much better for this.

Reflection can be encouraged or discouraged by design — this much is clear from the variety of internet-use helpers, like Moment and Intent. All of us (not just bullies and trolls) would use the Internet differently if we had more room for reflection.

Two lockscreens: one design encourages reflection, and one doesn’t. [from “Empowering Design”]

---

Exercise: On My Own Terms

In order to learn to support users in experimentation and reflection, designers must experiment and reflect on their own values. On My Own Terms is an exercise for this. Players fill out a worksheet, then socialize in an experimental way.

“On My Own Terms”. Join our community to play these games!

In the experimentation part, players defy norms they’ve previously obeyed, and see how it works out. Often they find that people like them better when they are less conventional — even when they are rude!

---

Here’s one thing this game makes clear: we discover what’s important to us in the context of real choices and their consequences. People often think they have certain values (eating kale, recycling, supporting the troops) but when they experiment and reflect on real choices, these values are discarded. They thought they believed in them, but only out of context.

This is how it was for me with consistency, rationality, masculinity, and being understated. When I played On My Own Terms, I decided to value these less.⁴ My true values are only clear through experimentation and reflection.

For users to have meaningful interactions and feel their time was well spent, they need to approach situations in a way they believe in. They need space to experiment and reflect.

But this is not enough.

Practice Spaces

Every social system makes some values easier to practice, and other values harder. Even with our values in order, a social environment can undermine our plans.

Most social platforms are designed in a way that encourages us to act against our values: less humbly, less honestly, less thoughtfully, and so on. Using these platforms while sticking to our values would mean constantly fighting their design. Unless we’re prepared for a fight, we’ll likely regret our choices.

There’s a way to address this, but it requires a radical change in how we design: we must reimagine social systems as practice spaces for the users’ values — as virtual places custom built to make it easier for the user to relate and to act in accord with their values.

Designers must get curious about two things:

1. When users want to relate according to a particular value, what is hard about doing that?
2. What is it about some social spaces that can make relating in this way easier?

For example, if an Instagram⁵ user valued being creative, being honest, or connecting adventurously, then designers would need to ask: what kinds of social environments make it easier to be creative, to be honest, or to connect adventurously? They could make a list of places where people find these things easier: camping trips, open-mics, writing groups, and so on.

Next, the designers would ask: which features of these environments make them good at this? For instance, when someone is trying to be creative, do mechanisms for showing relative status (like follower counts) help or hurt? How about when someone wants to connect adventurously? Or, with being creative, is this easier in a small group of close connections, or a large group of distant ones? And so on.

To take another example, if a News Feed user believes in being open-minded, designers would ask which social environments make this easier. Having made such a list, they would look for common features. Perhaps it’s easier to be open-minded when you remember something you respect about a person’s previous views. Or, perhaps it’s easier when you can tell if the person is in a thoughtful mood by reading their body language. Is open-mindedness more natural when those speaking have to explicitly yield time for others to respond? Designers would have to find out.⁶

---

Exercise: Space Jam

To start thinking this way, it’s best if designers focus first on values which they themselves have trouble practicing. In this game, Space Jam, each player shares something they’d like to practice, some way of interacting. Then everyone brainstorms, imagining practice spaces (both online and offline) which could make this easier.

“Space Jam”. Join our community to play these games!

Here’s an example of the game, played over Skype with three designers from Facebook:

Eva says she wants to practice “changing the subject when a conversation seems like a dead end.”

Someone comments that Facebook threads are especially bad at this. We set a timer for three minutes and brainstorm on our own. Then everyone presents one real-world way to practice, and one mediated way.

George’s idea involves a timer. When it rings, everyone says “this conversation doesn’t meet my need for ____”. Jennifer suggests something else: putting a bowl in the middle of a conversation. Player can write out alternate topics and put them in the bowl in a conspicuous but non-interrupting way. (Jennifer also applies this idea to Facebook comments, where the bowl is replaced by a sidebar.)

We all wonder together: could it ever be “okay” for people to say things like “this conversation doesn’t meet my need for ____”? Under what circumstances is this safe to say?

This leads to new ideas.

---

In the story above, Eva is an honest person. But that doesn’t mean it’s always easy to be honest. She struggles to be honest when she wants to change the conversation. By changing the social rules, we can make it easier for her to live according to her values.

Games like Space Jam show how much influence the rules of social spaces have over us, and how easy it is for thoughtful design to change those rules. Designers become more aware of the values around them and why they can be difficult to practice. They feel more responsible for the spaces they are creating. (Not just the spaces they make for users, but also in daily interactions with their colleagues). This gives them a fresh approach to design.

If designers learn this skill, they can support the broad diversity of users’ values. Then users will no longer have to fight the software to practice their values.

Sharing Wisdom

I hope the previous ideas—reflection, experimentation, and practice spaces—have given a sense for how to support meaningful actions. Let’s turn to the question of meaningful information and meaningful conversation.

We are having a problem in this area, too.

Amidst nonstop communication — a torrent of articles, videos, and posts — there is still a kind of conversation that people are starved for, because our platforms aren’t built for it.

When this type of conversation — which I’ll call sharing wisdom — is missing, people feel that no one understands or cares about what’s important to them. People feel their values are unheeded, unrecognized, and impossible to rally around.

As we’ll see, this situation is easy to exploit, and the media and fake news ecosystems have taken advantage. By looking at how this exploitation works, we can see how conversations become ideological and polarized, and how elections are manipulated.

But first, what do I mean by sharing wisdom?

Social conversation is often understood as telling stories, sharing feelings, or getting advice. But each of these can be seen as a way to discover values.

When we ask our friends for advice — if you look carefully — we aren’t often asking about what we should do. Instead, we’re asking them about what’s important in our situation. We’re asking for values which might be new to us. Humans constantly ask each other “what’s important?” — in a spouse, in a wine, in a programming language.

I’ll call this kind of conversation (both the questions and the answers) wisdom.

Wisdom, n. Information about another person’s hard-earned, personal values — what, through experimentation and reflection, they’ve come to believe is important for living.

Wisdom is what’s exchanged when best friends discuss their relationships or jobs, when we listen to stories told by grandmothers, church pastors, startup advisors, and so on.

It comes in many forms: mentorship, texts, rituals, games. We seek it naturally, and in normal conditions it is abundant.

For various reasons, the platforms are better for sharing other things (links, recommendations, family news) than for asking each other what’s important. So, on internet platforms, wisdom gets drowned out by other forms of discourse:

• By ideology. Our personal values are easily eclipsed by ideological values (for instance, by values designed to promote business, a certain elite, or one side in a political fight). This is happening when posts about partisan politics make us lose track of our shared (or sharable) concerns, or when articles about productivity outpace our deeper life questions.
• By scientism. Sometimes “hard data” or pseudo-scientific “models” are used to justify things that would be more appropriately understood as values. For instance, when neuroscience research is used to justify a style of leadership, our discourse about values suffers.
• By bullshit. Many other kinds of social information can drown out wisdom. This includes various kinds of self-promotion; it includes celebrities giving advice for which they have no special experience; it includes news. Information that looks like wisdom can make it harder to locate actual, hard-earned wisdom.

For all these reasons, talk about personal values tends to evaporate from the social platforms, which is why people feel isolated. They don’t sense that their personal values are being understood.

In this state, it’s easy for sites like Breitbart, Huffington Post, Buzzfeed, or even Russia Today to capitalize on our feeling of disconnection. These networks leverage the difficulty of sharing wisdom, and the ease of sharing links. They make a person feel like they are sharing a personal value (like living in a rural town or supporting women), when actually they are sharing headlines that twist that value into a political and ideological tool.

---

Exercise: Value Sharing Circle

For designers to get clear about what wisdom sounds like, it can be helpful to have a value sharing circle. Each person shares one value which they have lived up to on the day they are playing, and one which they haven’t. Here’s a transcript from one of these circles:

There are twelve of us, seated for dinner. We eat in silence for what feels like a long time. Then, someone begins to speak. It’s Otto. He says he works at a cemetery. At 6am this morning, they called him. They needed him to carry a coffin during a funeral service. No one else could do it. So, he went. Otto says he lived up to his values of showing up and being reliable. But — he says — he was distracted during the service. He’s not sure he did a good job. He worries about the people who were mourning, whether they noticed his missteps, whether his lack of presence made the ritual less perfect for them. So, he didn’t live up his values of supporting the sense of ritual and honoring the dead.

In the course of such an evening, participants are exposed to values they’ve never thought about. That night, other people spoke of their attempts to be ready for adventure, be a vulnerable leader, and make parenthood an adventure.⁷

---

Playing this makes the difference between true personal values and ideologies very clear. Notice how different these values are from the values of business. No one in the circle was particularly concerned with productivity, efficiency, or socio-economic status. No one was even concerned with happiness!

Social platforms could make it much easier to share our personal values (like small town living) directly, and to acknowledge one another and rally around them, without turning them into ideologies or articles.

This would do more to heal politics and media than any “fake news” initiative. To do it, designers will need to know what this kind of conversation sounds like, how to encourage it, and how to avoid drowning it out.

The Hardest Challenge

I’ve pointed out many challenges, but left out the big one. 😕

Only people with a particular mindset can do this type of design. It takes a new kind of empathy.

Empathy can mean understanding someone’s goals, or understanding someone’s feelings.⁸ And these are important.

But to build on these concepts — experimentation, reflection, wisdom, and practice spaces— a designer needs to see the experimental part of a person, the reflective part, the person’s desire for (and capacity for) wisdom, and what the person is practicing.⁹

As with other types of empathy, learning this means growing as a person.

Why? Well, just as it’s hard to see others’ feelings when we repress our own, or hard to listen to another person’s grand ambitions unless we are comfortable with ours... it’s hard to get familiar with another person’s values unless we are first cozy with our own, and with all the conflicts we have about them.

This is why the exercises I’ve listed (and others, which I didn’t have space to include) are so important. Spreading this new kind of empathy is a huge cultural challenge.

But it’s the only way forward for tech.

---

Thanks for reading. (Here are the credits and footnotes.)

Please clap for this and the previous post!

And…

• There is a community you can join to practice this stuff.
• Learn about a complimentary approach in metrics.
• Read more about human values, the connection between values and emotions, or the history of time well spent.

Making your own Face Recognition System

Making your own Face Recognition System

Hardik Gandhi   January 18, 2018

Face recognition is the latest trend when it comes to user authentication. Apple recently launched their new iPhone X which uses Face ID to authenticate users. OnePlus 5 is getting the Face Unlock feature from theOnePlus 5T soon. And Baidu is using face recognition instead of ID cards to allow their employees to enter their offices. These applications may seem like magic to a lot of people. But in this article we aim to demystify the subject by teaching you how to make your own simplified version of a face recognition system in Python.

Github link for those who do not like reading and only want the code

Background

Before we get into the details of the implementation I want to discuss the details of FaceNet. Which is the network we will be using in our system.

FaceNet

FaceNet is a neural network that learns a mapping from face images to a compact Euclidean space where distances correspond to a measure of face similarity. That is to say, the more similar two face images are the lesser the distance between them.

Triplet Loss

FaceNet uses a distinct loss method called Triplet Loss to calculate loss. Triplet Loss minimises the distance between an anchor and a positive, images that contain same identity, and maximises the distance between the anchor and a negative, images that contain different identities.

Figure 1: The Triplet Loss equation

• f(a) refers to the output encoding of the anchor
• f(p) refers to the output encoding of the positive
• f(n) refers to the output encoding of the negative
• alpha is a constant used to make sure that the network does not try to optimise towards f(a) - f(p) = f(a) - f(n) = 0.
• […]+ is equal to max(0, sum)

Siamese Networks

Figure 2: An example of a Siamese network that uses images of faces as input and outputs a 128 number encoding of the image. Source: Coursera

FaceNet is a Siamese Network. A Siamese Network is a type of neural network architecture that learns how to differentiate between two inputs. This allows them to learn which images are similar and which are not. These images could be contain faces.

Siamese networks consist of two identical neural networks, each with the same exact weights. First, each network take one of the two input images as input. Then, the outputs of the last layers of each network are sent to a function that determines whether the images contain the same identity.

In FaceNet, this is done by calculating the distance between the two outputs.

Implementation

Now that we have clarified the theory, we can jump straight into the implementation.

In our implementation we’re going to be using Keras and Tensorflow. Additionally, we’re using two utility files that we got from deeplearning.ai’s repo to abstract all interactions with the FaceNet network.:

• fr_utils.py contains functions to feed images to the network and getting the encoding of images
• inception_blocks_v2.py contains functions to prepare and compile the FaceNet network

Compiling the FaceNet network

The first thing we have to do is compile the FaceNet network so that we can use it for our face recognition system.

import os
import glob
import numpy as np
import cv2
import tensorflow as tf
from fr_utils import *
from inception_blocks_v2 import *
from keras import backend as K

K.set_image_data_format('channels_first')

FRmodel = faceRecoModel(input_shape=(3, 96, 96))

def triplet_loss(y_true, y_pred, alpha = 0.3):
    anchor, positive, negative = y_pred[0], y_pred[1], y_pred[2]

    pos_dist = tf.reduce_sum(tf.square(tf.subtract(anchor,
               positive)), axis=-1)
    neg_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, 
               negative)), axis=-1)
    basic_loss = tf.add(tf.subtract(pos_dist, neg_dist), alpha)
    loss = tf.reduce_sum(tf.maximum(basic_loss, 0.0))
   
    return loss

FRmodel.compile(optimizer = 'adam', loss = triplet_loss, metrics = ['accuracy'])
load_weights_from_FaceNet(FRmodel)

We’ll start by initialising our network with an input shape of (3, 96, 96). That means that the Red-Green-Blue (RGB) channels are the first dimension of the image volume fed to the network. And that all images that are fed to the network must be 96x96 pixel images.

Next we’ll define the Triplet Loss function. The function in the code snippet above follows the definition of the Triplet Loss equation that we defined in the previous section.

If you are unfamiliar with any of the Tensorflow functions used to perform the calculation, I’d recommend reading the documentation (for which I have added links to for each function) as it will improve your understanding of the code. But comparing the function to the equation in Figure 1 should be enough.

Once we have our loss function, we can compile our face recognition model using Keras. And we’ll use the Adam optimizer to minimise the loss calculated by the Triplet Loss function.

Preparing a Database

Now that we have compiled FaceNet, we are going to prepare a database of individuals we want our system to recognise. We are going to use all the images contained in our images directory for our database of individuals.

NOTE: We are only going to use one image of each individual in our implementation. The reason is that the FaceNet network is powerful enough to only need one image of an individual to recognise them!

def prepare_database():
    database = {}

    for file in glob.glob("images/*"):
        identity = os.path.splitext(os.path.basename(file))[0]
        database[identity] = img_path_to_encoding(file, FRmodel)

    return database

For each image, we will convert the image data to an encoding of 128 float numbers. We do this by calling the function img_path_to_encoding. The function takes in a path to an image and feeds the image to our face recognition network. Then, it returns the output from the network, which happens to be the encoding of the image.

Once we have added the encoding for each image to our database, our system can finally start recognising individuals!

Recognising a Face

As discussed in the Background section, FaceNet is trained to minimise the distance between images of the same individual and maximise the distance between images of different individuals. Our implementation uses this information to determine which individual the new image fed to our system is most likely to be.

def who_is_it(image, database, model):
    encoding = img_to_encoding(image, model)
    
    min_dist = 100
    identity = None
    
    # Loop over the database dictionary's names and encodings.
    for (name, db_enc) in database.items():
        dist = np.linalg.norm(db_enc - encoding)

        print('distance for %s is %s' %(name, dist))

        if dist < min_dist:
            min_dist = dist
            identity = name
    
    if min_dist > 0.52:
        return None
    else:
        return identity

The function above feeds the new image into a utility function called img_to_encoding. The function processes an image using FaceNet and returns the encoding of the image. Now that we have the encoding we can find the individual that the image most likely belongs to.

To find the individual, we go through our database and calculate the distance between our new image and each individual in the database. The individual with the lowest distance to the new image is then chosen as the most likely candidate.

Finally, we must determine whether the candidate image and the new image contain the same person or not. Since by the end of our loop we have only determined the most likely individual. This is where the following code snippet comes into play.

if min_dist > 0.52:
    return None
else:
    return identity

• If the distance is above 0.52, then we determine that the individual in the new image does not exist in our database.
• But, if the distance is equal to or below 0.52, then we determine they are the same individual!

Now the tricky part here is that the value 0.52 was achieved through trial-and-error on my behalf for my specific dataset. The best value might be much lower or slightly higher and it will depend on your implementation and data. I recommend trying out different values and see what fits your system best!

Building a System using Face Recognition

Now that we know the details on how we recognise a person using a face recognition algorithm, we can start having some fun with it.

In the Github repository I linked to at the beginning of this article is a demo that uses a laptop’s webcam to feed video frames to our face recognition algorithm. Once the algorithm recognises an individual in the frame, the demo plays an audio message that welcomes the user using the name of their image in the database. Figure 3 shows an example of the demo in action.

Figure 3: An image captured at the exact moment when the network recognised the individual in the image. The name of the image in the database was “skuli.jpg” so the audio message played was “Welcome skuli, have a nice day!”

Conclusion

By now you should be familiar with how face recognition systems work and how to make your own simplified face recognition system using a pre-trained version of the FaceNet network in python!

If you want to play around with the demonstration in the Github repository and add images of people you know then go ahead and fork the repository.

Have some fun with the demonstration and impress all your friends with your awesome knowledge of face recognition!