Видео

Thanks man ♥️ every comment helps!

Thanks for the suggestion, I’ll definitely put it on my list of ideas! I’ll have to be honest though, this is a bigger undertaking. You very likely have to use deep reinforcement learning (or something equivalent) to make any sort of reasonable progress. I’ll have to wait a bit more before I take on these bigger projects. For example the map location detector for counter strike took ages to develop and train, and it got less views than a simple Mario party video I put together on a few evenings. I’m working full time and don’t really have the capacity to start any large projects right now… maybe some day :) I could probably spin up a simple rule based bot that makes reasonable choices, but I don’t think I’d be satisfied by that.

And more is on the way, I’m already working on the next project :)

I noticed a typo, volume inconsistencies and some background footage flickering… I saw it has only 4 views, 2 of which were from me, so I thought I’ll fix it quickly. Reupload is coming soon! Thanks for the support :)

@@ProgrammingWithJulius thanks for letting me know

@@BahavianBrave reupload is done, by the way :)

Thanks man ♥️

There’s more coming! Already working on the next video :)

Thank you :) already working on the next project!

Thanks man :) more is coming soon!

Heyo ✋

Already working on the next video, subscribe so you don’t miss it :)

I deliberately did not look at any solutions before I finished mine, because I wanted to figure it out myself. But I had a look after I had enough games done for 2 videos - and I guess I won on goomba wrangler 😬😈

It was a lot of fun to program, too 😬

Yup, I guess I’ll always have this accent 😬 and thanks!

Thanks! There’s more coming soon :)

Thanks :) more is coming soon!

Erwischt 😬♥️

Thanks man, I really appreciate it! I’m working on a Mario Party Jamboree video next, but it’s a lot more pain to implement because it’s running on real hardware. Subscribe to not miss it :)

Thanks! ♥️

Thanks so much man, and I’m glad you enjoyed it. There’s more content coming, I’m already done with 3 more Mario Party DS games and right now I’m trying to get MP Jamboree to work 😬

it did for me.

Thanks! It seems a few more views are coming in now, let’s see how long that lasts :) but I’ve got some more work ahead of me

Thanks! Now it got a bit better at least :)

Thanks man, you too! Let’s see who cracks the algorithm first 😬

Thank you! :)

Thanks man! If you have some time, I would recommend this video to you: ruclips.net/video/V_xro1bcAuA/видео.html It's long, but it contains everything you need to get started with PyTorch. And whatever questions you still have, GPT (4, not mini!) can help with a lot of things, too. It's a lot of fun and really not that difficult to get into. I recommend starting with an image classifier AI, it's very simple and easy to see good results :) I made one too, in my last video: ruclips.net/video/RmUy7bBO_dE/видео.html Good luck!

Great idea, but sadly I don't have a fitting dataset... I would need a massive amount of demos. With that, you could go even further and predict, well, basically anything about the game

Thanks man, it means a lot ♥️ and I hope the subs are coming at some point, because this was a TON of work 😂

Much appreciated! More is definitely coming, and sooner.

I played the game for almost 4k hours, but I'm not actively playing premiere at the moment. And thanks! <3

Great catch, you had me scared I screwed something up in the dataset. Turns out I just screwed up when making the video. Here’s the real prediction result: imgur.com/a/pprq5gp

I almost can’t believe that - will definitely try it out. The work is done on the GPU and I’m not sure if that would be affected by privilege checks. I’m on vacation now though, so it will have to wait a while.

@@ProgrammingWithJulius well i remember i was running some python program in vscode and then in console, and then in console with administrator privileges, i think its just able to use more pc resources or more optimized? not sure if it will affect GPU part, but probably the CPU part

If you have any specific questions, I'll try my best to answer them :)

@@ProgrammingWithJulius too far over my head :( keep it up lmao

Thanks! More videos are definitely coming soon :)

Yes, all run data is saved in wandb.ai and the training graph animation was created with a little tool I wrote that takes the csv export and makes it into a video :)

And of course I can run the model on any specific image to get the predicted location

Thank you!

Thanks 🔥 I appreciate it!

Glad to hear, thank you!

Yeah, and the open source community has made it so easy to get into it. Only a couple years ago, the barrier to entry was much higher. I love how every prototype feels just a couple of lines of code away these days.

And thanks, of course :D

Thank you for your kind words! <3 Sadly the views already died down, it will probably end up at ~1500 views in a couple of days and then be buried forever. But I'll keep uploading, maybe one day :)

Thank you, I really appreciate it!

Thank you very much! If you know anyone who might be interested, feel free to share it. Every view counts for small channels like mine.

Thanks! :)

Well, the data doesn’t „need“ to be in that range. It’s just much easier to train for that. Imagine you’re an archer trying to hit a target. Now move that target 1000m up and away from you. Technically possible to still hit it? Yes, if you tape the arrow onto a rocket. Practical? No. Training for long enough, the neural net might eventually also learn to predict the real locations. But it takes much longer to tune the weights because they need to become absolutely massive to produce these numbers. There are more reasons why it’s impractical, like training stability and compatibility with activation functions, but this is the short version. Hope this makes it a bit clearer :)

Did you check out the website? I didn't add any analytics because I didn't want to deal with the cookie/imprint garbage...^^

And thanks! <3

Thank you so much for your kind words. I really appreciate it <3

Thanks for the sub and the kind words, I really appreciate it :) I will definitely upload more in the near future!

Thanks! <3 It was a really interesting project, albeit a lot more work than anticipated.

Thank you so much for your kind words! I'm hoping for it. Until then, I'll keep uploading :)

Thanks! I was worried as well, because it was one of these projects where you only know wether it works after you put in a lot of work. But I was really surprised at the results. 96 units off is basically spot on

Yeah, this channel is going to absolutely blow up in like 6 months.

Thank you so much ♥️