Hi I am new in ML/Ai Deep learning and I will star...

# kotlindl

[dhia chemingui]

Hi I am new in ML/Ai Deep learning and I will start with kotlin and computer vision any recommendation ?

[Peter]

If you want to build some front-end or integration to an existing AI model, Kotlin is a great choice. For example: https://blog.jetbrains.com/kotlin/2022/01/object-detection-with-kotlindl-and-ktor/

But if you want to define and train your own vision ML models, I’m not so sure. Most JVM/Java/Kotlin initiatives that target this use case, seem to get little attention 😢.

[Michal Harakal]

@dhia chemingui do you have particular use cases in mind?

[dhia chemingui]

Hi, yes, I am thinking about object detection and scene understanding with Ar . Honestly I never try ai .
and I want to start practice in this field.

[Michal Harakal]

If you're just getting into ML/AI and want to work with Kotlin + computer vision, a great way to really understand things is to build them yourself. One of my favorite ideas is based on Richard Feynman’s approach: “What I cannot create, I do not understand.”
Re-implementing core pieces of an ML framework— even small ones — is an amazing learning path. It might sound like a big undertaking, but since you're already here in the Kotlin community, it probably means you're interested in how things actually work under the hood. That aligns really well with the kind of work we’re doing with SKaiNET.
If you enjoy going deeper, try implementing the basics in Kotlin yourself: tensor ops, tiny autograd, a minimal convnet, etc. You’ll learn more from that than from any tutorial, and you might even find yourself contributing to existing Kotlin ML projects along the way.
Happy to help you get started if you want pointers — always great to have new people jump in!
Its still very early stages, but already having basic MPL on start and CNNs are just the current work in progress....

[Michal Harakal]

Regarding your use case "obejct detection", yolo8 is is currently also very high in the prioty pipeline ...

[dhia chemingui]

Yes , I really need some one like you .
The image is blurred for me but I like to forward with some pointers to learn and implement .
I am welcome to your guidance . Thank you .