How do I learn how to use a new framework?

Posted by Least_Letter_949@reddit | learnprogramming | View on Reddit | 9 comments

I'm trying to learn how to create an image detection LLM. I'm struggling with going from knowing 0 about a resource or framework to being able to code comfortably with it. I'm trying to learn how to use YOLO, but I'm confused about where to find resources about how to build an image detection model.

Experienced programmers,
What do you do when you are learning a new framework/resource?

Note: Please don't suggest AI. I'm trying to learn all of these things without the help of AI, so I better learn how to use the resource. I also don't want to rely on AI as a crutch.

Thanks!

[-]

illuminarias@reddit

Read documentation, look at existing projects that utilize that framework/library/resource, then experiment.

[-]

AdmirableBasil3154@reddit

pretty much what top comment said but i'd add that youtube tutorials are goldmine for yolo specifically. there's tons of step by step walkthroughs that show you actual coding process

also github is your friend - search for yolo projects and just read through the code even if you dont understand everything at first. you'll start recognizing patterns after looking at few different implementations

one thing that helped me when learning new stuff is starting with smallest possible project first. like maybe try detecting just one object type before going for complex multi-class detection

[-]

AutoNateAI@reddit

Yup, this is how literally all of us who were pre-AI learned (plus schooling, but even with schooling, we still did this…and still do).

I would suggest finding a YouTube video (no longer than 10mins) that explains object detection. You’ll realize you do not need an LLM for this. That would be a tool an LLM invoked for more context. But the tool itself, is a simple model.

You’ll realize that the LLM can help explain the concepts better, and honestly where you are, you should be using the LLM to learn more efficiently, not build rn.

It kinda sucks, but I’d say build these from scratch, and the only time you tap in with an LLM is to explain a concept that was introduced in a YouTube video. Once you get comfortable with building object detection algos, then that’s when you start offloading specific parts of the development (the parts you know well and know AI can handle) to your agent.

Hope this helps

[-]

Least_Letter_949@reddit (OP)

I never use it to build, just to explain concepts. I don't want to do that anymore because I feel I'm relying on it too much, though.

Thank you both for the replies!