I have a calibrated camera with extrinsincs and intrinsics parameters such that I can project a visible point in the world back to the image.
But I need to be able to deal with points that are also outside of what the camera can see.
how can I know if a point is visible by the camera? That would solve the issue when dealing with only one point.
how can I project a possibly only partially visible 3d box (8 points) back to the image?
My knowledge is limited and I cannot use a 3rd party framework to this end, rather need to implement it in python/C.
Related
I use godot to create my 3d game. I ran into a problem while creating portals using camera viewport rendering to texture. The problem is that the camera captures unnecessary objects that are behind portal. I partially solved this problem by setting the parameter "near " for the camera at a distance from the camera itself to the portal, but the part behind the portal began to be cut off.
The question is, is it possible to hide objects for a particular camera so that other cameras can see them? Perhaps there is another way to do this, for example by creating a static clipping plane?
Proximity Fade
Probably not what you are looking for, but I'll mention it for completeness sake.
The default material has proximity fade and distance fade, which you can use to make the material disappear if it is too close or to distant from the camera, respectively.
It is important to note that this is not a cull plane, and that the fading is gradual.
Thus, using proximity fade you can make objects near the camera appear semitransparent.
Using Visibility layers and cull mask
is it possible to hide objects for a particular camera so that other cameras can see them?
Every VisualInstance (you know, all things that are visible in 3D) has layers. And every Camera has a cull_mask. If the cull_mask of the Camera does not include any of the layers of a VisualInstance, then the Camera does not see that VisualInstance.
A VisualInstance with no layers will not show on no Camera, even if the Camera has all the layers in its cull_mask (which is the default).
You can either edit the cull_mask of the camera to not include the layers of the VisualInstance, or edit the layers of the VisualInstance, or both.
Using a custom shader cull plane
Perhaps there is another way to do this, for example by creating a static clipping plane?
You can use a custom spatial shader to cut things out based on a plane.
You need to define the plane as a uniforms. For this answer I'll use a point-normal definition of a plane:
n·(r - r_0)
That is:
dot(plane_normal, (world_position - plane_point)
Thus, we define a plane_normal and plane_point uniforms:
uniform vec3 plane_normal;
uniform vec3 plane_point;
The plane_normal gives us the orientation of the plane, while the plane_point is a point on the plane which allows us to position it.
And then use this logic:
vec3 wold_position = (CAMERA_MATRIX * vec4(VERTEX, 1.0)).xyz;
ALPHA = clamp(sign(dot(plane_normal, wold_position - plane_point)), 0.0, 1.0);
Here we are converting the coordinates of the current point to world space, and then using definition of the plane to find the points on one side (using sign), and set ALPHA based on that, such that everything on one side of the plane becomes invisible.
Note: This is not the only way to define the plane. Another popular definition is a 4D vector, where the xyz are the normal, and the w is the distance from plane to the origin.
Sadly, I don't think there is a way to make this work with multiple material passes, because ALPHA controls the blending of the passes, and will not result in transparency. And no, using discard; does not solve it either, because the other passes can write the fragment regardless. Thus, you are going to need to modify your materials to include that.
Further Sadly Godot 3.x does not support global uniforms (see Godot 4.0 gets global and per-instance shader uniforms). Which means you will have to set these parameter everywhere you need them.
Using Constructive Solid Geometry (CSG)
Add a CSGCombiner make the geometry that needs to disappear with other CSG nodes as children.
Then you can, for example, add a CSGSphere with operation set to "Subtraction", and move it with the Camera (for this purpose, I suggest to add a RemoteTransform node as child to the Camera and set its remote path to the CSGSphere).
Of course, it does not have to be a CSGSphere, you can use any CSG nodes for this purpose. For the portal, I imagine you could use a CSGBox and align it to the portal plane.
Note: Currently on Godot 3.3 CSG nodes do not support baking lights. This is a regression. See: Unable to bake lightmap with CSG due to the lack of ability to generate UV2 for CSG nodes.
Portals, actually
Bartleby Lawnjelly has a portal (godot-lportal) module for Godot 3.x.
Being a module, they require to build Godot from source. See Compiling on the official Godot documentation. It is not that bad, I promise. Or use build from godot-titan.
I have to explain that these portals are not portals in the Valve Portal video game series sense… The module lets you define areas as "rooms", and planes as "portals" that connect those rooms, in such way that you can look from one to the other. The purpose of this is to cull entire rooms unless you are looking through one of the portals.
Hopefully that makes more sense with a video. This is a somewhat old one, but good to get the idea across: Portal rendering module in Godot 3.2 - Improved performance. Seeing shadow pooping in the video? Bartleby Lawnjelly also has a custom lightmapper.
I placed a standard Camera in front of a moving Actor. When I set the current view to this camera I noticed a strange behaviour: If the actor get really close to another object on the scenario (a default cube) it disappear from the view. It looks like the camera is getting into the cube. I'm pretty sure the camera is not colliding with the cube because the actor has a couple of bumpers that prevent the side where the camera is placed to collide with other objects and the whole camera mesh is placed fully 'inside' the actor. The problem maybe is related with the size of the actor that's about 40cm x 30cm x 10cm. The observed cube is 1mt x 1mt x 1mt, the minimum distance of camera from cube is around 3 cm.
Sounds to me like you're experiencing an issue with an object passing your camera's "clipping plane." In the 3D world, this is simply just draw distance minimum and maximum values. For more information on what you are experiencing, check out this brilliant explanation by Autodesk: https://knowledge.autodesk.com/support/maya/learn-explore/caas/CloudHelp/cloudhelp/2018/ENU/Maya-Rendering/files/GUID-D69C23DA-ECFB-4D95-82F5-81118ED41C95-htm.html
Now, let's fix the issue! In Unreal Engine, it's super easy. Go into your Project Settings/General Settings. There is a value called Near Clip Plane, which simply changes the minimum clipping value for Camera components. I would bet making this value smaller will fix your issue! For a visual representation, check out this tutorial by Kyle Dail: https://www.youtube.com/watch?v=oO79qxNnOfU
I am just a beginner in game developing. Right now I am developing a game using Apple's Sprite kit and found out that the best way to position nodes on the scene is to provide percentages of width and height of the window boundaries as it makes sure all the nodes maintain their positions almost regardless of change in device display size. Using pixels to position nodes is not a peculiar idea as with the change in device display size of iPhone, nodes are either cut off or the scene squeezes leaving empty space around the scene boundary. I have watched how Apple recommend using scene editor but my issue is, using scene editor it allows you to position nodes by choosing pixels and not relative to to scene width or height. Am I making a mistake in understanding the scene editor capability. If I position all my nodes using scene editor as its saves a lot of time, how can I avoid problem with different iPhone sizes. I appreciate your help.
This is an age old problem, from all media formats.
You must decide, personally, what your favourite and most desirable target device is, and then make choices best for both it, and your creative process.
After making that decision you'll have to make your own decisions on how compromised you become on other devices, or how much you compromise your creative and production processes to benefit other device sizes and aspect ratios.
It's a balancing act.
And I strongly suggest favouring your favourite device and putting off all consideration of adaptation to other devices until after you've made something great.
Others will disagree.
i developed a game for the iphone4. Now i got problems with the iphone5 and the 4inch screen. My game is on the left side of the 4inch screen and i have a big black border on the right side. But the buttons from the game are in the middle of screen, they have same position like on the iphone4. I checked everythin but i dont know why the background-images and the sprites are on the left side and the buttons are in the middle. I want that everything is in the middle or on the left side. It would be great if anybody could help me!! Thanks!!
COCOS2d-iPhone:
If you're using the latest beta, the only change you should need to
make is export all your images at twice the size and use the "-hd"
suffix, similar to Apple's "#2x". The documentation also says you need
to set the content scale factor of the director.
You can find the documentation here.
and more detail here.
COCS2d-X:
Cocos2D-x has a very easy solution for multi-resolution problem.
In fact you just need to set your DesignResolution and then just imagine your target device will has this resolution.
If target device really has this resolution ( or some other but with same ratio) cocos2d will handle to fix screen and your game looks same in all devices.
And when ratio of target device is different, you have many option ( as cocos2d language, policy) to manage that.
For example if you use Exact fit policy, cocos2d will force your game ( and design) to fit target device screen (without that black boarder).
Exact fit
The entire application is visible in the specified area without trying
to preserve the original aspect ratio. Distortion can occur, and the
application may appear stretched or compressed.
For more detail just take a look at this link : http://www.cocos2d-x.org/projects/cocos2d-x/wiki/Multi_resolution_support
Why should you ever want something like this?
I want to track a single user that is mounted above the ground in a horizontal position. The user is facing downwards to allow free movement of legs and arms. Think of swimming for example.
I mounted the Kinect at the ceiling facing downwards so I have a free view of all extremities.
The sensor is rotated 90° in z-axis to have the maximum resolution (you're usually taller than wide).
Therefore the user is seen from the backside, rotated by 90°. It is impossible to get a proper skeleton from OpenNI 1.5. My tests showed that OpenNI is expecting the user facing the camera with the head up in y-axis (see my other answer). Microsofts SDK is the same but I excluded it here because it won't allow you to change the source code and cannot be adapted. OpenNI 2.0 is not working with the current SensorKinect to interface the Kinect in Linux. So:
Which class is generating the skeleton in OpenNI 1.5.x?
My best guess would be to rotate the prototype skeleton by y 180° and z 90°. If you know where I could find this.
EDIT: As I just learned there is no open source software that generates a skeleton from depth images so I fall back to the question in the header:
How can I get a user skeleton from a rotated back view?