So, my question that what function shoud I use, what is the fastest, or which is the most prevalent solution.
The solution alternatives:
-the vertex array
-polygon by polygon
-or another alternative
An object has triangles, quads and polygons which has more indices. But the vertex array I know can store only triangles, quads, or something like that. But how can I render the polygons? Is it right that I only render polygon by polygon the polygons wich has more indices? (I've heard the glMultiDrawElements, but it has similar data structure if I render poly by poly, isn't it?)
Or there's a good alternative I don't know?
You should convert the polygons to triangles and use a vertex buffer object.
Related
Basically I want a way to export the uvs from one model to another as a part of our pipeline where rig and texture/lookdev models(created simultaneously) need to be merged
I would like a solution other than importing the model into the scene and copying the uvset.
Something like an xml export.
Is there any way.
Thanks in advance
You can't really do this precisely unless the meshes are topologically identical. You can do a decent, but not perfect job with something like this:
for each triangle in the source model, derive a tangent space matrix. That's the matrix which converts the world space points of that triangle into the UV points of that face.
for each triangle in the target model, see if it matches one of the triangles in the source. "Match" will mean "has the same three corners" regardless of order (assuming you can fix up your normals in a separate step
Where source triangles have exact matches in the target triangles, apply a planar project that matches the tangent space matrix.
where a target triangle doesn't have an exact match you'll have to guess; you can try finding the N closest matches and doing a weighted blend between all of their matrices or something like that - but it will be hacky
This should be visually pretty close to the source, but all of the UV triangles will be disconnected; you'll need to merge coincident UVs to prevent seams.
It's a pretty non-trivial project, unfortunately.
I am trying to do skeletal animation in legacy OpenGL and thought I could use matrices on individual vertices. When I programmed it and it didn't work, I did some Googling to find this: https://www.talisman.org/opengl-1.1/Reference/glLoadMatrix.html
GL_INVALID_OPERATION is generated if glLoadMatrix is executed between the execution of glBegin and the corresponding execution of glEnd.
So now I'm stumped. Here is a diagram:
Bones are labeled in red. I'm trying to do skeletal animation so there are two rectangles. One uses Bone 0 and the second uses Bone 1. Only specific vertices of the triangles that make the second rectangle use the rotation matrix of Bone 1, and the ones that don't use the rotation matrix of Bone 0, kind of making a snake, if that makes sense.
Since I cannot use glLoadMatrix for individual vertices in a triangle, what other way can I displace a vertex based on a stored matrix? Perhaps multiply some of the matrix values to the vertex? Not sure how to go about doing that. Any input is appreciated, thanks!
You mention the two rectangles Bone0 and Bone1. You need to draw them separately since they need to have separate transformation matrices. Two points of the two rectangles are coincident, your transformation matrix to draw Bone1 must ensure that:
glTranslateF(...);
glRotateF(...); /* position rectangle Bone0 */
glBegin(GL_QUADS); /* draw rectangle Bone0 */
glVertex3f(...); /* draw it */
...
glEnd();
glPushMatrix(); /* save transformation matrix */
glMulMatrix(...); /*
* as per your drawing, this is not just a
* simple translate/rotate operation, but
* a translate/shear
* you need to do that manually
*/
glBegin(GL_QUADS); /*
* draw rectangle Bone1. Two of the vertices are
* coincident with two of rectangle Bone0. Your
* shear matrix must ensure they are
*/
glVertex3F(...);
glEnd();
What you're trying to do is called skinning! And unfortunately it will involve a bit more effort than your approach. It is possible to do it between one begin and end, which is generally preferable.
The easiest way is not to use OpenGL to transform your vertices. Use your favourite matrix math library to multiply the vertices with your bone matrices before they get passed to OpenGL. If the number of vertices is not too large, it won't slow you down much.
The harder way is to implement a skinning shader. This book chapter provides a good introduction on how that is done. The principle is to upload multiple matrices to OpenGL, and give each vertex an index which says which matrix to multiply with. This will be much faster than the easy approach.
GPUs are fast because they are optimized for doing the same operation on a large set of data - the tradeoff for this is that you can't modify the state (such as changing the matrix) while a draw call is in progress.
Heh, have been doing a lot of math stuff since I posted my question and checked back just now to post an answer for anyone with the same question, and I noticed I already have some answers, so thanks for answering with your input!
Since I have figured out a solution, though, I thought I would post an answer along with these other two.
Basically, what I am doing with the rendering is per-vertex rendering anyway, where it reads the vertices of each triangle from a data buffer and all of that, so it wasn't too much trouble to go ahead and write a custom function to multiply a matrix to a vertex, so a copy of the vertex is loaded from the buffer, the matrix is multiplied to it based on which bone the vertex is mapped to, and then that is used for rendering that particular vertex on the triangle.
Funny that I already had implemented what #Hannesh suggested and I just had to write the multiplier function. Very cool!
The harder way is to implement a skinning shader. This book chapter provides a good introduction on how that is done. The principle is to upload multiple matrices to OpenGL, and give each vertex an index which says which matrix to multiply with. This will be much faster than the easy approach.
Thanks again! I'll up-vote whenever I have the reputation to do so!
Within a leaflet polygon object, there are two arrays "_originalPoints" and "_parts", I was wondering if anyone knew what the purpose of these two arrays were.
Thanks,
I'm Leaflet author.
_originalPoints is an array of projected geographical points (screen coordinates of the polygon points), and _parts is an array of arrays of points that eventually gets rendered after clipping and simplifying points. In case of polygon different "parts" are needed to render holes in polygons, and in case of polylines you often get several paths out of one after clipping (cutting points off a certain rectangular area for performance).
Im trying to triangulate a polygon for use in a 3d model. When i try using the ear method on a polygon with points as dotted below, i get triangles where the red lines are. Since there are no other points inside these triangles this is probably correct. But i want it to triangulate the area inside the black lines only. Anyone know of any algorithms that will do this?
There are many algorithms to triangulate a polygon that do not need partitioning into monotone polygons first. One is described in my textbook Computational Geometry in C, which has code associated with it that can be freely downloaded from that link (in C or in Java).
You must first have the points in order corresponding to a boundary traversal. My code assumes counterclockwise, but of course that is easy to change. See also the Wikipedia article. Perhaps that is your problem, that you don't have the boundary points consistently organized?
The usual approach would be to split your simple polygon into monotone polygon using trapezoid decomposition and then triangulate the monotone polygons.
The first part can be achieved with a sweep line algorithm. And speed-ups are possible with the right data-structure (e.g. doubly connected edge list). The best description of this, that I know, can be found in Computational Geometry. This and this also seem helpful.
Wikipedia suggest that you break the polygon up into monotone polygons. You check that the polygon is not concave by simply checking for all angles being less than 180 degrees - any corners which has a angle of over 180 is concave, and you need to break it at that corner.
If you can use C++, you can use CGAL and in particular the example given here that can triangulate a set of non-intersected polygons. This example works only if you already know the black segments.
You need to use the EarClipping algorithm, not the Delaunay. See the following white paper: http://www.geometrictools.com/Documentation/TriangulationByEarClipping.pdf
I had a generalized question to find out if it was possible or not to do matrix calculations on a rectangle. I have a CvRect that has information stored in it with coordinates and I have a cvMat that has transformational data. What I would like to know is if there was a way to get the Rect to use the matrix data to generate a rotated, skewed, and repositioned rectangle out of it. I've searched online, but I was only able to get information on image transforms.
Thanks in advance for the help.
No, this is not possible. cv::Rect is also not capable of that, as it only describes rectangles in a Manhattan world. There is cv::RotatedRect, but this also does not handle skewing.
You can, however, feed the corner points of your rectangle to cv::transform:
http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=transform#cv2.transform
You will then obtain four points that are transformed accordingly. Note that there are also more specialized versions of this function, e.g. warpPerspective() and warpAffine().