Need ideas for a project

[Nineshadow]

So I have some free time these days and I figured it would be best if I used it for something. I want to start a small project and see what I can end up with. I'm not really sure what to do though; I'm out of ideas.

My language of choice is C++ so that's what I'm going with.

 

A game is always a possibility. I've done those before, but tbh, the biggest issue I have with them is getting artistic content : sprites,models,etc. Drawing really isn't my thing. I can do 3D modelling, but not too well. But if I were to make a game, I'd certainly try a Metroidvania. Anyway, some time ago I went through openGL and got myself used to it. I have some experience in graphics programming. I could maybe use my previous work to complete a game engine? I also have experience in embedding Lua for use as a scripting language, so there's little work to do to implement that. So it would leave me with the core structure of the game engine itself and the physics engine, probably.

 

I've also though about making a prime number generator (for use as a benchmark maybe?) and optimizing it as much as I can. There's definitely quite a lot to be done in that department : wheel factorization, dividing the workload in pieces small enough to fit into the registries, parallelism, etc. It doesn't sound too exciting though.

 

So...let's do some brainstorming; just come up with something. I would be grateful for any  piece of creativity.

[trag1c]

Since you already have experience with graphics programming you could do life like terrain with fractal algorithms such as diamond square. It can create some pretty interesting stuff. Ocean simulation as well is pretty interesting.

[Hazy125]

AI. Machine learning algorithms. You can apply and test on twitch chat for example. Try to develop and algorithm that detects trends and comment popularity so your bot doesn't accidentally ban(fake ban in your program) people saying 'lol' or some such. Or don't use twitch. Try to scrape webpages and use your AI to correctly identify useful information and store it in a database. From there you can try and figure out a useful and efficient way to search the database. Or don't do any of that, my creativity is zero. But developing an AI for something is always an interesting project. I just don't have any time to do something like that myself

 

 

EDIT: Oh, since you've experimented with game stuff, maybe make a better fish AI. Current fish AI is meh

[Nineshadow]

2 minutes ago, Hazy125 said:

AI. Machine learning algorithms. You can apply and test on twitch chat for example. Try to develop and algorithm that detects trends and comment popularity so your bot doesn't accidentally ban(fake ban in your program) people saying 'lol' or some such. Or don't use twitch. Try to scrape webpages and use your AI to correctly identify useful information and store it in a database. From there you can try and figure out a useful and efficient way to search the database. Or don't do any of that, my creativity is zero. But developing an AI for something is always an interesting project. I just don't have any time to do something like that myself

Hmm...I actually wanted to get into deep learning some months ago. Read a little bit and it seemed interesting, but forgot about it the next day.

[cluelessgenius]

i always wanted to do something with body tracking via camera but ont have the time right now.

[WiiManic]

1 hour ago, Nineshadow said:

Hmm...I actually wanted to get into deep learning some months ago. Read a little bit and it seemed interesting, but forgot about it the next day.

Deep Learning was going to be my suggestion. There are so many random datasets out there for you to mess around with and test different forms of networks on.

I've just completed my dissertation on the automated recognition of blood vessels using neural networks and its been incredibly interesting.

There is so much going on in the area and the ideas and its such an interesting thing to be working with.

 

Its not C++ but a decent library and with decent amounts of support and tutorials would be Theano (or there are other libraries built on top of Theano).

Though using something like that would require some amount of knowledge in Python.

I found it incredibly fun to use and powerful too, and seemingly very quick since the majority of the back-end can be ran using CUDA very easily. 

Also helped Uni gave me access to a Tesla K40c to run it all on.

 

If not Deep Learning, my other interest would be something based around Genetic Algorithms since they are very fun to mess with.

I've less experience implementing them, so I've not got a recommendation for anywhere to look unfortunately, but I've seen a lot of fun things done with them.

Something that learns to play games, or something that learns to perform a task, it really depends on what you want it to do. 

[Nineshadow]

1 hour ago, WiiManic said:

Deep Learning was going to be my suggestion. There are so many random datasets out there for you to mess around with and test different forms of networks on.

I've just completed my dissertation on the automated recognition of blood vessels using neural networks and its been incredibly interesting.

There is so much going on in the area and the ideas and its such an interesting thing to be working with.

 

Its not C++ but a decent library and with decent amounts of support and tutorials would be Theano (or there are other libraries built on top of Theano).

Though using something like that would require some amount of knowledge in Python.

I found it incredibly fun to use and powerful too, and seemingly very quick since the majority of the back-end can be ran using CUDA very easily. 

Also helped Uni gave me access to a Tesla K40c to run it all on.

 

If not Deep Learning, my other interest would be something based around Genetic Algorithms since they are very fun to mess with.

I've less experience implementing them, so I've not got a recommendation for anywhere to look unfortunately, but I've seen a lot of fun things done with them.

Something that learns to play games, or something that learns to perform a task, it really depends on what you want it to do. 

Yeah, looks like a good idea. I know Python pretty well, just haven't really used it for anything.

Do you have any good resources for deep learning by any chance?

[WiiManic]

11 minutes ago, Nineshadow said:

Yeah, looks like a good idea. I know Python pretty well, just haven't really used it for anything.

Do you have any good resources for deep learning by any chance?

Honestly, that Theano tutorial I linked is a decent start, its got a bunch of code examples for network types on the MNIST dataset.

I started with reading and trying all of the ones in that, well the interesting and relevant ones. 

It will give you an idea of how the network functions at the start in the theoretical way, before the description of the code to say how.

If it were me I'd start with that and maybe run through some of the shorter examples (some of them can get to hours long, even with CUDA acceleration).

Then you'll have a decent understanding of how the networks and Theano work, then maybe look into some of the other libraries built on top of Theano if you implement your own networks. Implementing your own networks is easy enough directly in Theano once you've got the loading routine altered, its just a little more awkward than some other libraries that allow you to load images with a single function.

[rm -rf]

3 hours ago, Nineshadow said:

So I have some free time these days and I figured it would be best if I used it for something. I want to start a small project and see what I can end up with. I'm not really sure what to do though; I'm out of ideas.

My language of choice is C++ so that's what I'm going with.

 

A game is always a possibility. I've done those before, but tbh, the biggest issue I have with them is getting artistic content : sprites,models,etc. Drawing really isn't my thing. I can do 3D modelling, but not too well. But if I were to make a game, I'd certainly try a Metroidvania. Anyway, some time ago I went through openGL and got myself used to it. I have some experience in graphics programming. I could maybe use my previous work to complete a game engine? I also have experience in embedding Lua for use as a scripting language, so there's little work to do to implement that. So it would leave me with the core structure of the game engine itself and the physics engine, probably.

 

I've also though about making a prime number generator (for use as a benchmark maybe?) and optimizing it as much as I can. There's definitely quite a lot to be done in that department : wheel factorization, dividing the workload in pieces small enough to fit into the registries, parallelism, etc. It doesn't sound too exciting though.

 

So...let's do some brainstorming; just come up with something. I would be grateful for any  piece of creativity.

A game where you are on a Möbius strip, that could be an interesting mechanic

[mathijs727]

My project for the coming summer is learning either DX11 or DX12 (already know OpenGL).

Probably gonna make a render engine as thats what interests me most.

Hopefully I can include some pretty advanced stuff in my engine like voxel global illumination, screen space reflections, ambient occlusion (already done SSAO in OpenGL), automatic shadow mapping + prebaked shadows.

 

If you're going with deep learning then there are a lot of good nearal network libraries like googles Tensorflow.

It has GPU acceleration but its Nvidia (CUDA) only

[vorticalbox]

4 hours ago, Hazy125 said:

AI. Machine learning algorithms. You can apply and test on twitch chat for example. Try to develop and algorithm that detects trends and comment popularity so your bot doesn't accidentally ban(fake ban in your program) people saying 'lol' or some such. Or don't use twitch. Try to scrape webpages and use your AI to correctly identify useful information and store it in a database. From there you can try and figure out a useful and efficient way to search the database. Or don't do any of that, my creativity is zero. But developing an AI for something is always an interesting project. I just don't have any time to do something like that myself

 

 

EDIT: Oh, since you've experimented with game stuff, maybe make a better fish AI. Current fish AI is meh

+1 on the fish AI. i want fish to react to sounds and movement and food ^_^. Vbox needs a virtual fish tank that reacts to my mouse.

[Azgoth 2]

So when I was an undergraduate, one of my final projects (in a computational physics class) had to do with acoustics and finding resonant frequencies of arbitrarily shaped spaces.  You could do something like that, e.g. make a model of a room or something, put a sound source in it, and make that sound propagate according to the wave equation (protip: I used a simple central difference numerical method, which is pretty accurate with small time steps and relatively simple).  Then you could model acoustics of rooms or arbitrary spaces, e.g. for doing realistic sound work in a game or something.

 

Or fractals.  Something like Mandelbrot or the Burning Ship or a Newton Fractal is always a fun thing to get out.

[prolemur]

19 hours ago, Hazy125 said:

EDIT: Oh, since you've experimented with game stuff, maybe make a better fish AI. Current fish AI is meh

110% agree, we need to be pouring more money into fish ai