So there are a ton of answers so forgive me if I repeat much of what has already been said, but here is my 2 cents.
First pick an idea. Any idea will be fine, something simple would probably be better then large. Projects have a tendency to grow in their scope very quickly (some call it feature creep).
Next make a skeleton for the project. This will require a little bit of architecture and design knowledge and you'll probably get it wrong the first ten times you try it - I did. Simply lay out a decent file structure and maybe a small skeleton of code that shows the important parts of the system.
Save the skeleton in your VCS (pick your poison with this one and hold on when it leads to a holy war). Once you've started using VCS, constantly using it for small changes becomes second nature, but make sure to start.
Now pick a small, but important feature for the system and make it. Don't worry about making sure you have everything encapsulated perfectly and that it has the "best" design (that will evolve with the system). Just get something that will work. Also getting some unit tests will help ensure you know what happened when something breaks, if you run the tests regularly.
Build your system. This would also be a good time to get an automated build system and continuous integration. If you don't know what they are then either learn it and try, or just continue at your own risk; either way keep working.
Now pick another feature and repeat and repeat and repeat.
Once you think it works well, show it to a friend. The friend doesn't have to know how to program or even know what the program does. One you'll feel good about showing to someone and two it will help you know exactly what the system does.
If you get to the point where you are very confident with what you made, release it online and try and get feedback. A repository hub or the programmers sub-reditt might provide you with some constructive criticism. Try and find a CS/SE professor and have him/her look at it. Maybe ask a professional programmer. Just get another programmers opinion.
Once you finish (or probably before) you'll realize that the code you initially wrote is a lot worse than what you made recently. That is perfectly natural and happens to all of us. You now need to find a new project and learn something new - maybe a new testing strategy or how to use Service Oriented Architecture.