[FREE] GPU Line of Sight / Field of View

Since I no longer have the time to maintain this asset myself, I've decided to open source it and make it available on GitHub!






A line of sight system visualizes which parts of the game world can be seen from the standpoint of for example a 3rd person in-game character. The area outside of the line of sight is obscured and remains hidden from the player.

This system was developed as an alternative to CPU based line of sight visualizations. CPU based systems use a combination of ray casting and dynamic meshes to compute the line of sight and are often CPU intensive, causing a serious impact on performance.

This new GPU based line of sight system uses a technique very similar to shadow mapping and is significantly faster than any CPU based system, freeing up valuable CPU time. It has been designed from the ground up with performance, ease-of-use and customizability in mind. The effect can be tweaked to achieve any desired visual style. The user can control the look of the effect by selecting which image effects are applied to the area outside of the line of sight. It is perfectly suited for top-down stealth or action games, but also works from other viewpoints and in other type of games.

This system is compatible with both the forward and deferred rendering pipelines and supports DX11. Because this system makes heavy use of render textures, it is currently not supported on mobile platforms.

Now available on the GitHub!

Try out the demo!

Main Features:

• Fast: designed with performance in mind, zero allocation during run-time
• Fully customizable effect: match the style of your project seamlessly
• Works with both perspective and orthographic cameras
• Includes scripts to check if objects are inside line of sight
• Includes full documentation + example scene

Supports:

• Forward and deferred rendering pipelines
• DX11

Not Supported:

• Sprite or 2D toolkit based projects
• Mobile and WebGL platforms

* Works with Unity 5 Personal Edition
* Requires Pro license for Unity 4.x

 

Nice. I wish Unity Free had the render textures. Besides line of sight, which is really a lighting/shadowing thing, do you have other types of lights like point lights etc?

 

What do you mean exactly by point lights?

If you mean a 360 degree line of sight, it's very easy to create using this system.

 

Ahh I see, cool.

 

Hi there!

It looks really nice! =) How complex is the process of setting up the scene and FOV´s?
Is it possible to combine the FOV of different agents together?
Does your solution handle 3d environments / height variation?

Thanks in advance

 

Thanks!

Setting up the scene and getting the system up and running is really easy and should only take a few minutes.

This system technically allows an unlimited amount of agents / line of sight sources (you can add more agents in the webplayer demo by pressing space bar). Even low spec systems should be able to handle more than 10 agents on the screen at once (my PC can handle a 100+). As an optimization, agents which are off screen are also not rendered.

Yes, 3D environments with height variations work. It will also work from for example a first person perspective.

 

Great top hear that!
The visual effect is separated from the FOV calculation itself? Would it be possible to access the raw data from the "visible" objects while avoiding any visual output?
PS: It´s planned to work on skinned meshes as well?
Sorry for the amount of questions!

 

Yes, the visual effect is separated from the calculations. The system basically let's you chose which image effects to apply to the area that's outside of the line of sight, giving you complete freedom to change the style.

You could avoid any visual output by simply not adding any image effects. I'm not sure why you would want to do that, but it would work

I'm not sure what you mean by raw data, but this system includes a script component that you can attach to an object to check if it's inside the line of sight. The "NPCs" walking around in the webplayer demo use this script to hide themselves when not inside the line of sight.

This effect works on skinned meshes work without a problem

 

This asset was approved and is now available on the Asset Store!

Full link: www.assetstore.unity3d.com/#/content/19764

 

This asset looks pretty nice, I've looking for something like this for weeks, my game use a isometric cam not a top down cam, do you think this effect would works for me?

 

Yes, this effect can work from any perspective, it doesn't require your game to be top down.

Here's a quick example of this effect being used with an isometric camera and 2 line of sight sources.

 

Here's a list of things I want to add in the next version:

• Add color groups: line of sight cones can be rendered in different colors
• Add invert mask option: apply effects to area inside of line of sight
  
• Improve bias accuracy

If you guys have any suggestions or feature requests for the next version, be sure to let me know!

 

Mobile support would be essential in my case and is what is holding me on buying this.
I was looking for accessing the raw data so that I can use this for both FOV and LOF without the visual feedback itself.
Cheers

 

I would love to support mobile in the future. I'm pretty sure this version should technically run on mobile, but I'll make a test build for android to make sure. Because this system makes heavy use of render textures, I'm just not sure about its performance on older mobile devices which makes it impractical to support.

That being said, this system doesn't require render textures, to check if an object is inside the line of sight. So if you don't want any of the visuals, but just need to know if an object is visible, this could definitely run on mobile.
But this system was primarily designed for the visual effect, so using it just for that might not be the most efficient or practical solution.

 

Humm, got it!
If I manage to get some free time I´ll buy and work on it in the next week! =) Thanks a lot for all the answers.
Btw, you have an amazing portifolio!

 

Glad I could help, and thanks for the compliments on my portfolio

 

After making some changes I've managed to get my line of sight system working on Android, so mobile support will probably be included it in the next update as well!

I still need to do more testing to see if I can get the performance to an acceptable level on mobile devices, but I'm hopeful.

 

Great to hear that! =D

 

I rewrote a big part of my system to make it compatible with mobile devices, but it still has some major issues I need to solve. Most of the issues are related to Unity bugs I need to circumvent, so they are rather tricky.
Because of this I will probably delay mobile support until version 1.2.

For version 1.1 I'll be adding these features:

• Ability to render Line of Sight sources in different colors
• HDR mode (more info soon)
• Some fixes and optimizations

I hope to be able to submit version 1.1 this weekend, so ti should be available some time next week!

 

Btw, let me ask: Do you need to use a camera for each FOV calculation? In that case does rendering all those extra cameras generate an overhead on render passes / poly count?
I ´ll buy it this weekend. Great work

 

Yes, you need a camera for every line of sight source, but this camera only renders a part of the scene using a fast depth only replacement shader.

So adding an extra line of sight source adds a bit of overhead, but only if the source is visible on screen. You can minimize this overhead by keeping the camera's (attached to the source) far clipping plane as small as possible.

 

Hi there! I just bought GPU LoS =)
This asset is great! I managed to do a fast test integrating it with Manufactura K4 Top-Down Interiors. Here´s my early result:

In this case I didn´t excluded the furniture from the calculation process, so I´ll have better results later.

One thing that I still didn´t figured out is that I´m getting some artifacts on walls at the demo scene:
Edit: Following pdf manual, antialias solved it.


Great work on this asset! =D I´m eager to see what will be available next.
Cheers!

 

Hi there!
I´ve managed to get some interesting results for a 360* FOV using a single camera.

In this case the camera is over the character and is pointing down. I had to adjust far clipping plane until it reaches this circular volume. As this requires the camera to use a wide FOV, it messes with frustum calculation and raycasts but it´s worth considering this approach for some specific situations. In that case a distance check might handle the false positive.

 

Hi Tycho,

I'm considering purchasing this asset, but I'm still not sure if it will be useful for my type/genre of game. I'm building a real-time strategy game (think StarCraft 2 or Command and Conquer style) that is "third person" top-down basically.

All I'm looking for is a fast way to calculate whether or not a ranged unit can see its enemy target without anything blocking its line of attack (such as a building between the unit/agent and it's enemy target or a large hill).

I need ranged units to be able to detect if they do not have line of sight to their target, and if they don't, then re-path to their enemy target until they do. I also would need to integrate the line of sight checks with my own scripts, so is there a Boolean variable or something that would say "yes this unit has line of sight" that I could then relay to my own script?

I'm going to have about 150 units for each player max, so each game might have about 300 total units maximum. Of course, not all of them will be on screen at the same time. A typical scenario would be about 40 - 60 units on screen at one time. Would your system still be peformant under these use-cases?

Thanks for any info.

 

Glad you like the asset and thanks for the purchase! The screenshot with the manufactura assets looks really promising!

I've found another solution to the artifact problem you show in your second screenshot, which doesn't require enabling anti-aliasing. If you set the clear flags property of the first camera to depth only, this will also fix the issue without any performance impact. This won't allow you to set a clear color though, but that will probably not be a problem for a top down game.

The proper way to do a 360 degree line of sight, would be to combine 4 cameras, each with an FOV of 90 degrees. This way the frustum calculations still work correctly, and tweaking the line of sight cone becomes a lot easier! Here's a quick screenshot showing my technique:



I'll make sure to include a 360 prefab in the next update (which I'm planning to submit today).

 

To be honest, I don't think this system would be very suitable for your needs.

It does include a script to check if an object is inside the line of sight (which sets a boolean value you can check from other scripts), but it does that for all line of sight sources / agents at the same time. This would make it impractical to check if one unit is inside the line of sight of another unit, and would require quite a few code changes.

I hope this helps!

 

Yep, I first tried using 4 cameras for the 360, but wanted to check if I could have better performance with a single one.
It will be really useful being able to render Line of Sight sources in different colors =)
I would recommend including a mask for the raycast, there can be situations where devs have an invisible wall that shouldn´t break LoS check.
Cheers

 

Concerning that subject, do you guys know how Warcraft 3 handled their Fog of war and FOV? It was very efficient. Never found any reference to the technical details of their magic =)

 

Different colors will be available in this update

Good point about the raycast mask, I'll check if I can still include it in this update, but if it's too much work it will be for 1.2

 

Yes that helps. Thank you. I appreciate you being honest about it

 

I've submitted version 1.1 and it's now awaiting approval!

Here's the changelog for version 1.1:

• Added HDR mode to LOS Mask
• Added Color Mask property to LOS Source
• Added Intensity property to LOS Source
• Added Out of Bound Area property to LOS Source
• Added Invert Mask option to LOS Source
• Added Layer Mask property to LOS Culler
• Added example prefabs
• Added LOS Layer Excluder script component
• Simplified setup process for multiple cameras
• Improved performance

Here's an image showing the effect of the HDR mask in combination with colored LOS sources:

 

I am planning to buy this asset once it reaches mobile optimization, still I was reading the forum and got an answer for Velo222

Velo222 just make a mesh collider for your units, if you can have the mesh with less than 100 tris it will be great to handle "OnTriggerStay" collisions

 

Version 1.1 is now live on the Asset Store!

 

I'm also in need of mobile support. I will purchase this over the weekend and take a look.

 

Hey Tycho, this looks great! I'm considering using this in a prototype I'm working on, but I have a few questions.

1) Is this suitable for lighting? I'd like to have dynamic lighting simulated in my game, and this looks perfect if the performance is good enough for 20+ lights at once.

2) I'd also like to use this for a field-of-view system. Can I layer this effect, once for lighting then again for player FOV?

Thanks

 

Hey Simie,

1: This system works with any kind of lighting ( static / dynamic, forward / deferred ). The amount of lights doesn't influence the performance of the LOS effect, so if it runs smoothly before, it should run fine with this effect enabled.

2: What do you mean exactly with layering?

 

Hi Tycho,

I was referring to using this effect as the actual lighting - ie the scene is rendered without any lighting (fullbright) and this effect simulates light sources.

By layering I meant that after that lighting pass of this effect is applied, can I then apply another pass for player vision?

 

Bump?

 

Hey I just picked up this asset is it possible to see if one of the players is actually visible in the LOS? For example I was working on a stealth game and I wanted to be to know if the enemy actually saw the player or not.

 

Hi there!
It´s worth taking a look at LOSCuller script for that.
Cheers

 

Ok I am also interested with this solution but I have few questions.

It mentions that it doesn't work with sprite or 2dtoolkit. I am guessing here, but I think it is because this tool requires the camera to render a depth map from the source and uses that to calculate / decide if a given pixel is seen or not. This raises few questions.

1. Is this mean it is technically possible to make it work with 2dtoolkit , if I provide some extra geometry that is only being rendered by the source camera that simulates the depth for the 2d sprites?

2. If it is actually using the depth map and does extra rendering using camera etc.. how is it different than using normal light source from Unity? Is it differ by performance? I can understand that it may detect the object inside FOV , but as far as underlying systems and performance goes , how does it compare to the other usual Unity rendering system? If I simulate light source by using this system, will it run faster than the Unity deferred lighting system? If so why?

Thanks advance for your reply!

 

Hello!

I would like to implement this kind of solution for a zombie game, instead of raycasts and colliders, is possible to not use the image effects stuff? this solution is aplicable to what i need? its a third person game, i would like to have precise character detection, and this looks great

Many thanks

 

I saw other assets like cone visions and somethings like that, but i found out that this one has GPU support,

 

I would like to buy this product, but first i have a question:

- Can i make it raycast 360 degrees all around the character instead of just making a narrow cone?

Also, in the isometric image you posted, is the blue cube floating? Cause im making an isometric game and the light seems to be below the cube, it looks wrong, as if the cube is floating.

 

Sorry for not replying to any of your questions, but for some reason I stopped getting notifications for this thread

Let me catch up by answering all your questions

 

Yes, it should be possible to mask a dark and light version of your scene using this system to simulate lighting and performance should be faster than using dynamic lights.

And yes, you can layer this effect.

 

Yes, your assumptions are correct.

1: Yes, that would be a possible work around, but might be tricky to set-up.
2: It is different from normal lighting, since I doesn't affect the lighting on objects, it just creates a mask which you can use to blend 2 different versions of your scene, for example the original one, and one altered using post processing or even rendered from a different camera.

 

Yes, you're not required to use image effects.

 

Yes, this asset includes a prefab which you can use for a 360 field of view.

And yes the blue cube is floating in that screenshot, and the light is indeed beneath it

 

Can i define a Y height as to where i want the raycast? What i mean is... How does the script choose what is a solid object or not? Does it raycast from the ground up, or just a specific height? A good example could be: what if i have a window at a certain height on the wall, and another at another height. How can i decide what should obstruct the view in this three dimensional case? Or, better yet, if i have a wall thats not too tall and i want it to obstruct the view, as its an isometric game, it might be chopped in half and still block it.

Also, i would like to know if it comes with that blur effect, along with the black and white. Can i choose just one or another? I dont want the light cones on stuff with line of sight, just the blur and maybe the black/white where they cant see. Cause the objects wont be light sources, ill have another light sources of my own. So they had to not emit any colors/light.

In your previous answer, is the cube at a greater height from the ground or the light is not matching the exact cube place? Its confusing.

Could you provide a much more complete DEMO with the options to tweak this kind of stuff and some that others have suggested too?