DEV Community: Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ

How does AI bounding box detection work?

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Fri, 13 Aug 2021 16:29:49 +0000

🧠 Learn in 30 seconds

It looks so simple when AI does it right?

But the AI doesn't give you an image, it gives you data. It's up to you to make it look simple. The results you get are presented in a way so you can choose specific items from all the returned results.

You might think a box gives you four values, and you're right, but it only gives you TWO points. From that you can infer a box to draw that identifies the object.

However... it very rarely gives you one box. Lots of times you get a whole bunch of boxes for you to chose from.

This is because the model is often trained to return X results no matter what, and sometimes it's meant to detect a whole wide array of objects, not just one.

Many boxes are identified by what they are detecting, like people or tables etc. But you might have a photo with three people in it, and it returns 30 people boxes. Each of these boxes generally give you a score of confidence, and it's your job to identify if they are identifying the same object, or separate objects.

There's some well known algorithms to take the best of the best.

Once you have the boxes you're interested in, you can draw them easily with a canvas. In JavaScript you just overlay a canvas on the image and draw rectangles there.

The final result lets you draw clean object detection on websites. It's pretty cool and easy to do this on images and video.

Want the code to do this? It's all in Chapter 6 of my book with a deep explanation.

I'm looking forward to what you can do with your very own AI object detection! Don't forget to tag your cool stuff with the #MadeWithTFJS Hashtag on Twitter!

And don't forget to follow me for more AI tidbits like this!

Creating TimeWarpScan.me

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Thu, 03 Jun 2021 19:01:57 +0000

You don't have to use AI tech like TensorFlow.js for AI only. I enjoyed adding GPU acceleration to my JavaScript to make a wild project. Want to know what it is? Watch the promo video I made:

Building with JavaScript is quite rewarding. I hope you get the feel of it, too. Just watch us enjoy the site on a Walkthrough Wednesday with friends.

https://www.twitch.tv/videos/1044068063

I've been fortunate enough to talk about this at Magnolia JS conf and other places! It's a fantastic Open Source project for people learning and looking to contribute during their #100DaysOfCode

The site is here: http://timewarpscan.me/
The source code is here: https://github.com/GantMan/timewarp/tree/main/timewarpscanme

The site is hosted on AWS Amplify and written in React!

And of course, if you want to be a master of TensorFlow.js to make your own fun projects, here's the obligatory book link:

https://amzn.to/3dR3vpY

AI & JavaScript with TensorFlow.js

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Tue, 27 Apr 2021 19:20:54 +0000

I've been having a blast with AI for years, and the pace keeps accelerating.

I can't wait for every code newbie to be able to build the future with AI and JS. Here's a lil blog post on my current vibe with ML in JavaScript.

I keep getting invited to conferences and meeting new people who are surprised that this tech is real. After one talk, they see how to apply TensorFlow.js in new ways and products.

The web applications are numerous!

ALSO: I'm proud to mention I'll be featured on a @ReactNativeRdio episode where I'll talk about the cool stuff in JS mobile dev. Episode dropping soon!
https://reactnativeradio.com/

It reminds me of that time on stage when I did AI in React Native live! I still smile every time I watch this video.

NicOrNot Stage Demo

The truth is people who code are unleashed when they learn what they can do. That's why we've lead Open Source in TensorFlow JS.

We've created libraries that improve quality of life for everyone, and made it free. NSFWJS is used around the world and has 4.5k stars.

https://nsfwjs.com/

But computer vision can be complex to a new Front-End dev, so it's also important to know when to be fun, too. Hopefully, there's enough Trekkie & Harry Potter Nerds to make my poorly drawn examples make sense
http://aisortinghat.com/

If you're into Web Development then I highly recommend you watch
@jason_mayes on twitter.
He runs the TensorFlow.js show and tell videos on YouTube.

Or at the very least subscribe to his #MadeWithTFJS hashtag and see your fellow devs create

Lastly, my book by O'Reilly Media is in Quality Check 2, which means you can pre-order it on Amazon, so you can invest in you and your team's education.

It won't be long before every JavaScript Developer is wielding Deep Learning.

Get the Book: https://amzn.to/3dR3vpY

Live-coding, Creating, Living

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Mon, 09 Nov 2020 14:08:36 +0000

Live-coding, Creating, Living

Give yourself the gift of creativity

We folks at Infinite Red are pretty darn good. I'd say we're one of the best React Native shops, if not the best in the world. But our strength comes from our willingness to grow, experiment, and communicate. Let me give you an example.

I was chatting with the epic Michael Chan from the React Podcast, and we laughed that we sometimes join meetings and then check our appearance. I'm guilty of fixing my hair quickly as soon as I see myself 😅 . Michael said that he knows someone who has an app that is basically a mirror. So they can check their appearance from their toolbar. SO SIMPLE!

But why buy an app? Can't we do this faster, cross-platform, and easily in a website? How long would it take to buy a domain, build a website, and deploy it for this EXACT need? As it turns out... less than 30 minutes of coding and 30 minutes to package and deploy. I live-coded the whole thing.

Fun right? If you don't watch the whole video, feel free to skim-through and get the basics of what's happening. It's pretty simple, but lots of fun! I've updated/upgraded the code since.

What did you update after this video was made?

If you watch the video you can see I have a mistake where the website selects the wrong camera. I've added a dropdown to the site that lets you pick your camera and the website remembers. For the most part, that's all!

How the hell do I share a screenshot of this site?

QuickMirror.me The site works great!

You can access the source and send awesome pull-requests here:
infinitered/quickmirrorme

OR, just try it out live! https://quickmirror.me/

So Gant, what's the lesson here? - You

We fell in love with Opens Source Software, because we code in public, we share our knowledge, we fail, we grow, and we build connections. I still love creating Open Source code, but I think it's time to evolve. This year the pandemic has demanded social distancing, but in effect, I feel distanced socially. I miss my fellow humans, and my Twitter feed is filled with those who miss me. So I think it's a good time to level-up sharing, caring, and coding.

Try to build your whims, ideas, and dreams. It's up to you to stream it, record it, or share it. The world needs more useless websites, neat ideas, and high-bandwidth videos. We can learn together. Let me know what you're building.

Gant Laborde is a co-owner and Chief Innovation Officer at Infinite Red, published author, adjunct professor, worldwide public speaker, and mad scientist in training. Tweet or visit him at a conference.

Oh Hey! Don't forget to share this article!

The Jetson Nano - Dirty Secrets

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Mon, 23 Dec 2019 17:11:15 +0000

🙀 JETSON WUH? WAT IZIT?

The NVIDIA Jetson Nano is an amazing little computer. I'd describe it as a lot like a Raspberry Pi, except it has a GPU. Combine that with it's low $99 USD price tag, and AI/ML enthusiasts see a cheap device that can go from desktop to edge with perfect parity. An ideal machine for an AI in the wild!

🎅 Correction: Make that $89 USD today!

https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-nano-developer-kit/

BONUS SELLING POINTS THAT GOT ME EXCITED

As a Mac user, I often find myself with a wimpy GPU. This little device might outperform my MacBook Pro for AI/ML training.
Some semi-important libraries, like Dlib, aren't built for Windows, so you need a Linux machine to get that GPU bump.
Any good excuse for running Linux again.

My Goal?

I have an old 2012 iMac that sits to the right of my station, that I don't generally use even though I could switch my mouse/keyboard with the press of a button.

Instead, I could replace this space with the Nano. BUT… It's not a crafting table; I'm going to need to close it up like little computer and keep it safe.

Something like this GeekWorm case will do!

Sounds like a plan! I can just order me a $30 case on top the $99, and I'm set!

… or am I?

Dirty Secret #1 - Price in Parts

You're not done buying. Time to load your cart up with a few extra items just to get the Nano running.

SD Card: Much like the Pi, the SD card is not included. Most of us are used to this for edge devices. Since I want a desktop-level device, I sprung for the 256GB micro card. +$35 USD

Power: The 5v 2a USB power supply can't handle peripherals. SURPRISE! You'll need to get a 5v 4a barrel jack so you don't lose your screen. +$14 USD

Fan: Some cases come with fans. The one I got did not. Throw a 5v fan in to cool the GPU. +$15 USD

Wifi + Bluetooth: Unlike the latest Pi devices, there's no WiFi or Bluetooth, so you'll have to buy that card separately. +$21 USD

Antenna: The WiFi card does not include antennas. I was crazy and attached an internal antenna inside on the first port, and with the second antenna, I ran it to the external of the case. +$14 USD

This sneaks the price of my desktop from $130 (device + case) up to $230.

TL:DR; While this modular system lets you keep the price down if you're embedding the Jetson Nano into a device for the field, it's a bit of a sticker shock from the original $99 price you might have been considering.

Dirty Secret #2 --- DIY Only

This device is perfect for the computer builder, the robot maker, and the all-out puzzle solver. If you've got a workshop, you'll be happy!

If you don't have a workshop or time, you're in trouble.

The screw holes for the fan are not tapped, the parts unpacked fill a small table, and the card installation requires a little know-how. It's like IKEA furniture. If this is your first time, it's going to take a while, and as soon as you're done, you'll reflect on how you could have done it in a quarter of that time.

While I've done some wild stuff with Raspberry Pi's, I'll admit I was looking up all kinds of conversions while I was at the hardware store, and ultimately over-buying just to assure I wasn't going to get the wrong size.

TL:DR; Nothing replaces that feeling you get when you dress the wires and screw the final screws, but it's important to note you're going to need to do all that yourself 👆 --- it's quite a barrier between purchase and use!

Dirty Secret #3--- You're Still Not Ready Yet?

Finally! I've got this little black box sitting on my desk, it's wired up, tested, and ready to go! I load up Ubuntu, and I go to install a package, and.... pip is not installed. Python is there, but quite a few common libraries you'd expect to be on the image are completely missing.

Now I get that you might not need everything... but for those of us who do, we've got a big install ahead of us, and given that it's a small CPU, we've got hours of installs.

Fortunately, here's a fantastic article that enumerates the next few hours of your life: https://www.pyimagesearch.com/2019/05/06/getting-started-with-the-nvidia-jetson-nano/

This tutorial doesn't get OpenCV running, so you'll need to build that from source. Hopefully, we'll find a good tutorial for that soon.

TL:DR; You can expect hours of setup before running your first project on the device.

The Dirty Summary

I guess the big issue is BTB (barriers to bench-marking). If you're looking to simply evaluate the Jetson Nano as a tool, you're biting off days of ancillary tasks and unknown costs to pit this device against an old laptop or phone.

The on-boarding isn't necessarily difficult, just significant. The playful whimsicality of "trying it out" like most people do a Raspberry Pi is absent unless your idea of trying it out is having the budget and time to tinker. This begs for someone to ship a "Getting Started" kit or pre-built enclosures with all the headache removed.

Maybe I'm betraying my developer/engineer psyche by appealing to a marketing "Do it for me" approach, but as more edge devices emerge, I believe onboarding time saved is customer interest earned.

Next Steps:

I'll be evaluating the Jetson Nano as a friendly second computer. Benchmarking it, playing with it, and hopefully writing more about the experience! Be sure to follow me on here and other social outlets for updates!

Gant Laborde is Chief Innovation Officer at Infinite Red, published author, adjunct professor, worldwide public speaker, and mad scientist in training. Clap/follow/tweet or visit him at a conference.

See more about Machine Learning here: FunMachineLearn Twitter and here https://infinite.red/machinelearning

Have a minute? Read more!

AI Demystified --- A Free Course\
*5-day certificate course to get a handle on AI buzzwords*shift.infinite.red

Perf Machine Learning on Rasp Pi\
*3 Frameworks for Machine Learning on the Raspberry Pi The revolution of AI is reaching new heights through new mediums...*www.freecodecamp.org

Looking to get into AI? Take this free course!

Infinite Red Academy\
*Chief Innovation Officer for Infinite Red Gant Laborde is an owner of Infinite Red, mentor, adjunct professor...*academy.infinite.red

Awesome TensorFlow.js - Rock Paper Scissors

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Tue, 06 Aug 2019 17:43:16 +0000

Machine Learning can be pretty complicated. Fortunately, we can always bring some entertaining aspects to counteract the mountain of linear algebra. I'm a fan of Machine Learning, which is why I'm all about @FunMachineLearn on Twitter. It's also why I'm starting a newsletter for awesome stuff in AI.

Recently I decided to tackle the complexities of Machine Learning making it as simple as pressing buttons in a browser. The results were awesome!

If you're interested in training your own AI in your browser to do Rock Paper or Scissors recognition, then you're going to enjoy my latest published article.

https://heartbeat.fritz.ai/using-tensorflow-js-to-train-a-rock-paper-scissors-model-b5f393b548eb

If you're interested in other crazy things you can do in Machine Learning in the browser, I highly recommend you check out my NSFWJS library.

https://nsfwjs.com/

Gant Laborde is Chief Innovation Officer at Infinite Red, a published author, adjunct professor, worldwide public speaker, and mad scientist in training. Clap/follow/tweet or visit him at a conference.

Building Something Fun - NSFW Detector

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Wed, 20 Mar 2019 19:07:16 +0000

Coming up with useful applications of Machine Learning can be an exciting and complicated problem. Sure, I can do the same linear regression problem everyone does, with the same data, but if I'm not excited about what I'm creating, how can I expect anyone else to be interested?

Then I got the idea, what about NSFW detection for client and server! This could cut down on work for teams, as well as provide a high-profile project for others.

Since creating the NSFW model, throngs of people have tweeted, and shown up to add pull requests. I'll even be Keynoting in Italy, to review the project.

One common request, a lot of people keep asking for the training data. I'm both entertained and engaged in all the interactions in this project. Never would I have assumed so many people would help me learn, challenge me to grow, and take interest in my project.

Because of this, I've documented the entire project here: https://medium.freecodecamp.org/how-to-set-up-nsfw-content-detection-with-machine-learning-229a9725829c

And released a JavaScript library here: https://shift.infinite.red/avoid-nightmares-nsfw-js-ab7b176978b1

Make something awesome

So next time you're handed the same old projects and the dusty roadmap of projects you should follow, just reject it; spice it up. Build something fun, crazy, and interesting. You'll be surprised how rewarding it can be.

Nic Cage or Not - React Native

Gant ℒ𝒶𝒷ℴ𝓇𝒹ℯ — Mon, 08 Oct 2018 22:22:34 +0000

Machine Learning is doing crazy new things, finding sharks, creating faces, and more wild and useful examples. For instance, there’s a Machine Learning model that could identify who would steal the declaration of independence. That means some crazy person spent time training a model to identify Nic Cage. From installing beta software to learning how to properly train an image classification model. Who would do such a thing? Spoiler alert, it was me.

While extremely useful, the Nic or Not app also has some interesting technology under the hood. Facial recognition will continue to advance and consequently more and more apps will have an interesting application of such. If you’re interested in the future of mobile and the fun of real-time facial recognition enjoy the story of creating the ‘Nic or Not’ App in React Native.