Pysource

If you’ve ever tried detecting tiny objects—like ants, cars in drone footage, or people in a crowded scene—you know how frustrating it can be.



Even with powerful object detection models like YOLO, the accuracy is often terrible in these scenarios.



Objects get missed, detections are inconsistent, and results are unreliable.



But what if you could:

✅ Detect tiny objects with high precision (even from a drone or satellite view)
✅ Accurately track people in crowded places
✅ Improve your object detection model without expensive hardware upgrades



I recently tested a method that dramatically improved detection accuracy in all these cases. The difference?

Massive.



I break it all down in this video:

2 months ago | [YT] | 16

Pysource

Last week, I decided to give ChatGPT o3-mini-high a shot to help me write some computer vision code.

I expected something useful—but what I got? Wow.

It didn’t just give me some half-baked snippet. It gave me a full, clean, readable, and actually usable piece of code. Exactly how I like it.

I plugged it in, and within minutes, I had my solution working. Since I got results way faster than expected, I kept going.

What started as a simple Python script turned into hundreds of lines of code across multiple files.

A week later? I had the foundation of a full project.

And now, I’m even considering turning it into a SaaS.

This got me thinking…

Programming is changing fast. We saw the first big shift when ChatGPT launched, and now, with tools like GitHub Copilot, Cursor, and more, AI-powered coding is the new reality.

So, will AI replace programmers?

Not anytime soon. But what will change is what makes a great programmer.

Here’s what I see happening in 2025 and beyond: 👇

🔹 You need deep foundational knowledge – Understanding data structures, project architecture, and how everything fits together is still essential. AI can assist, but you need to make the right decisions.

🔹 You need to master a framework – Even if AI writes most of the code, you must have the confidence to debug, adapt, and optimize when things don’t work as expected.

🔹You need to get good at AI prompting & adaptability – Knowing how to work with AI tools will be just as important as knowing how to code. The best developers will be the ones who can adapt quickly—and that includes you.

Productivity is skyrocketing—and so are expectations.

What do you think? Are developers ready for this shift?

2 months ago | [YT] | 21

Pysource

I’m currently working on a new project and a new video.

Every time I come up with an idea for a new project or video, I get that rush of excitement—you know the one.

But then two things usually happen:

1. I start overthinking it. I’ll find all the reasons not to do it—why it might fail, why I’m not ready, why it’s not the “right time.” And before I know it, the idea’s gone.

2. Or (and this is the better path), I just start. Even if I have no clue what I’m doing or where to begin, I’ll do something—anything—that pushes the idea forward.

For example:

If I’m thinking about building an advanced vehicle tracking and analysis system, I won’t start by trying to perfect it. I’ll just write a few lines of code to detect vehicles in an image.

If I want to create a web app, I’ll open Canva and sketch out the UI. Or I’ll write a tiny function that tackles the app’s main purpose.

These small actions create momentum. And that momentum? It’s a game changer. It will give you the energy, enthusiasm, and confidence to keep going, even if the final idea never makes it into the world.

So here’s my advice: if you have an idea, just start. It doesn’t matter how small or imperfect the first step is—what matters is that you take it.

Who knows where it might lead?

2 months ago | [YT] | 6

Pysource

Where can you apply computer vision?

Whenever I get asked about computer vision by someone who’s unfamiliar with it, I always give the same two examples:

Imagine a manufacturing company that needs to identify defects on a conveyor belt. By adding cameras and some smart algorithms, we can automate the entire visual inspection process.

Or take traffic analysis: you can count people, vehicles, or even types of vehicles using cameras to get insights for city planning or congestion management.

Everyone gets these examples—they’re relatable, they’re practical, and they show how computer vision can solve real-world problems.

But the truth is, computer vision can do so much more.

Think of it as a friend with super-sharp eyes who can see things, track data, and even give advice based on what it “sees.”

For example, a while back, I collaborated on a project with a PhD researcher from Bristol University. We built an app to help kids learn math in a more interactive and engaging way.

And recently, I’ve been toying with the idea of building something for myself. Last year, I wasn’t as consistent with my workouts as I wanted to be, so I thought: What if I built an app that tracks my bodyweight exercises at home? I could gamify it—make it fun and help myself stay committed. (If I ever do it, I’ll share it here!)

What about you? Have you worked on any cool computer vision projects?
I’d love to hear about them!

2 months ago | [YT] | 8

Pysource

Being an AI developer and freelancer can be incredibly rewarding—but let’s be honest, it can also be incredibly frustrating.

One day, you’re in “WOW” territory, building something innovative, surprising yourself with what you’re capable of. And then there are those moments when a client sees your work as pure wizardry—while you’re quietly thinking, “That wasn’t even that complex.”

But here’s the flip side: frustration. A lot of it.

If you’ve been doing this for a while, you’ve probably faced this scenario: a client brings you a shiny, exciting project, and you struggle to price it properly. Maybe you underestimate the complexity (happens to the best of us), and suddenly you’re deep in a project that’s far bigger than you anticipated.

And if you’re not careful, you end up working under a ton of pressure, doing something really challenging, without being compensated fairly.

I’ve made that mistake more times than I’d like to admit.

Here’s the lesson I wish I had learned earlier: don’t let money be your first driver. Of course, there are times when compromises are necessary—that’s just reality—but the best projects align with your values and passions.

When you find that alignment, it doesn’t feel like work anymore. It feels like playing a game, where each day is a new level to unlock and explore.

If what you’re working on right now doesn’t feel that way, it might be time to start searching. Change projects. Change clients.

The right work is out there—and when you find it, the difference is night and day.

I wish you a happy weekend

2 months ago | [YT] | 21

Pysource

Do you want to use AI Vision to build a visual inspection system?

Don't miss the latest video.

If your goal is to build a computer vision solution to identify in defects in real time, then today's video will give you an understanding of how to build such solution.

You don't need to be an expert neither you need to know computer vision. You need to have some python skills and basic of programming and be eager to put in the work.

With the technology and algorithms we have available today, so many things are possible that you'll be surprise what you can actually build if you're not afraid to experiment with new things.

Have you got any questions about it?

leave a comment

5 months ago | [YT] | 13

Pysource

Pysource Community is now live 🎉

Join before 19th of November to get access to a huge launch deal.

This is an academy for you if you're a developer, a freelancer or a startup founder and you want to build AI Vision solutions or you simply want to level up your computer vision skills.

If you're serious about computer vision and you want to get access to premium content, you don't want to miss this deal.

To know more watch the link in the comments

5 months ago | [YT] | 21

Pysource

Want to speed up Opencv and get 2 times frame rate?
I've been working on a code that runs Python with Opencv and multithreading to make this happen easily and with a few lines of code.

you can watch the full tutorial (latest video) if you want to speed up your computer vision project.

ps. Soon I'm also bringing a tutorial about speeding up YOLO and Object Detection up to 4x

6 months ago | [YT] | 48

Pysource

Hi there,

I've just released a new video about the Opencv AI Kit (OAK-D).

Let me know in the comments what projects you would like to see using this device.

4 years ago | [YT] | 22

Pysource

Hi Everyone,
I want to thank everyone who is commenting and giving me feedbacks about the new videos I'm releasing.

I'm happy that most of you are enjoying the Neural Network from scratch series.

A new video is out, and now we're at the 7th lesson of the video series. Let me know what you think about it.

https://www.youtube.com/watch?v=3zcHG...

4 years ago | [YT] | 12