Jump to content
  • Advertisement
Sign in to follow this  
knobby67

OpenGL md5 vanishing textures

This topic is 3346 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi all, I hope some one can help me out with this, I've just started working on loading MD5 models, all was going well until I tried to load textures, I've "flipped" the co-ordinates so they map on opengl co-ordinates. However when I run the animation parts of the texture keeps vanishing, so about 50% look right, 40% disappear/reappear as the animation runs, 10% flicker. I've got the co-ordiantes from the model which i've ran on a view and they work on that exe. I'm only loading in one texture, the diffuse map, but that shouldn't make any difference. Do I need to recalculate texture coordinates, like the meshes? Has anyone else experienced this, if so can you advise? Thanks.

Share this post


Link to post
Share on other sites
Advertisement
I played with md5 model loading not so long ago - how exactly do you load your texture coordinates - they should be directly loaded and used (I've seen no deformations of texture coordinates so far) and of course flipped in Y axis.
Aren't you computing them on run using weights (which is incorrect - but it might cause deformations)?
Or don't you have wrong animation (because coordinate moves with vertex)?

Share this post


Link to post
Share on other sites
Hi no I don't transform the texture coordinates on the fly. I just load them from the file.

However I do change the x,y,z of the vertexes so the correspond to the coordinate system of opengl (eg move y is up ect), I then flip the textures so they are look right, eg head texture maps onto head.

However I have found the vertext "flips" are slightly different to these I use on 3ds and AES models I flip these by

Vx=Vx
Vy=Vz
Vz=-Vy

with md5 I do

Vx=Vy
Vy=Vz
Vz=Vx

This "looks" right but perhaps is wrong?



Share this post


Link to post
Share on other sites
I'm not sure but it may be because you have set GL_CLAMP to your texture parameters. Inverting v coordinate just flipping its sign leads to an out of bound mapping clamped by OpenGL.

u = u
v = -v

Try instead

u = u
v = 1-v

Share this post


Link to post
Share on other sites
#EmmeKappaErre - this would cause wrong visiblity of texture even while not animating, and it'd probably do this with the texture even before starting the animation (anyway as I think from what I've read this is not a problem)

Although you're right - one should always use clamped coordinates (if it's not your intention to use unclamped onex).

#knobby67 - Yeah, this is common thing - that one with swapping vertices while loading. The axis stuff is probably causing the problems - let me post you some pseudo code of mine (lets say I've loaded the verts, weights and coords, quaternions, etc. as they are ... this code is very unoptimised, just to show off how do the stuff most simple way - you can try then some GPU skinning, etc. but It's complicated and not good for "showing the way"):
// first I need to transform all the joints:
// For every joint
for(int i = 0; i < numberOfJoints; i++)
{
// Get it's ID
md5.animations[animationNo].joints.id = i;
// and parent ID
md5.animations[animationNo].joints.parent = md5.animations[animationNo].parent;

// If it has no parent
if(md5.animations[animationNo].parent == -1)
{
// Set it's position and quaternion as the position and quaternion defined for current frame of animation
md5.animations[animationNo].joints.pos = md5.animations[animationNo].frames[currentFrame].joints.position;
md5.animations[animationNo].joints.quat = md5.animations[animationNo].frames[currentFrame].joints.quaternion;
}
// Otherwise it has some parent
else
{
// Get me position and quaternion of current joint from current frame of animation, though it needs to be transformed
// by parent joint (because it's in some kind of HIERARCHY)
tVector3 thisPosition = md5.animations[animationNo].frames[currentFrame].joints.position;
md5_quaternion thisQuaternion = md5.animations[animationNo].frames[currentFrame].joints.quaternion;

// Get me current joints parent joint into "parent" variable
md5_joint parent = md5.animations[animationNo].joints[md5.animations[animationNo].parent];

// Now lets compute some rotation
// First we need to rotate current joint by its parent joint rotation (e.g. rotate it with all hierarchy prior to it)
tVector3 rotatedPos;
rotatedPos = RotatePointAroundQuaternion(parent.quaternion, thisPosition);
// And now comes position computing - simply use position after rotating and add parents position (transforming it to
// right place - where the joint should be)
md5.animations[animationNo].joints.position.x = rotatedPos.x + parent.position.x;
md5.animations[animationNo].joints.position.y = rotatedPos.y + parent.position.y;
md5.animations[animationNo].joints.position.z = rotatedPos.z + parent.position.z;

// And at last some quaternion math - joints current quaternion must be computed as multiplication of its and the
// parents quaternion, of course normalized (for importance of this and why this is important see some quaternion
// math book)
md5.animations[animationNo].joints.quaternion = MultiplyQuaternions(parent.quaternion, thisQuaternion);
md5.animations[animationNo].joints.quaternion = NormalizeQuaternion(md5.animations[animationNumber].joints.quaternion);
}
}

// Now I need to compute right positions (I'm showing just computing position of first vertex in triangle, the other two are the same (except you need to use different vertex indicies)
// Lets do it the simple way ... legacy glBegin/glEnd forever :D ... just joking, almost everyone is using VBOs and so,
// so it's just to show it the simpliest way possible.
glBegin(GL_TRIANGLES);
// Cycling cycling - for every mesh and every triangle in it (showing just first vertex in triangle)
for(int i = 0; i < md5.numMeshes; i++)
{
for(int j = 0; j < md5.meshes.numTri; j++)
{
// Texture coordinates 1.0f - coordinateY isn't necessary, even -coordinateY will work, but to be sure about
// clamping stuff
glTexCoord2f(md5.meshes.vert[md5.meshes.tri[j].a].texCoord.x,
1.0f - md5.meshes.vert[md5.meshes.tri[j].a].texCoord.y);
// tVector3 is some kind of structure that holds 3 floats - X Y and Z
tVector3 finalVert;
// Initialize it (I init is defaultly as this, but to show it needs to equal (0,0,0)
finalVert = tVector3(0.0f, 0.0f, 0.0f);
// now lets sum the weights to get correct position
for(int k = md5.meshes.vert[md5.meshes.tri[j].a].weightBegin; k < md5.meshes.vert[md5.meshes.tri[j].a].weightBegin + md5.meshes.vert[md5.meshes.tri[j].a].weightNum; k++)
{
// Gimme weight and joint for this weight
md5_weight weight = md5.meshes.weight[k];
md5_joint joint = md5.animations[animationNumber].thisFrame.joints[weight.jointId];

// And now some stuff with rotation (weight needs to be rotated around joints quaternion)
tVector3 temp = tVector3(0.0f, 0.0f, 0.0f);
temp = RotatePoint(joint.quat, weight.pos);

// And finally add it's position to vertex (based upon the weight "effect" on it
finalVert.x += (joint.pos.x + temp.x) * weight.weightStrength;
finalVert.y += (joint.pos.y + temp.y) * weight.weightStrength;
finalVert.z += (joint.pos.z + temp.z) * weight.weightStrength;
}
// The sum of all that weight computing into finalVert gives us correct position of vertex, and just right now you flip
// X Y Z into X Z-Y ... that's all
glVertex3f(finalVert.x, finalVert.z,-finalVert.y);

// Now you need to do the same for other two vertices of triangle
}
}
// And thy mighty procedure to complete this code-show and fulfill the purpose of my life :D
glEnd();






I appologize for long post

Share this post


Link to post
Share on other sites
Thanks for the help everyone, and thanks for the code example, but I think that's basically what I'm doing? Except I change my texture when I load using
the texture matrix

glMatrixMode (GL_TEXTURE);
glLoadIdentity ();
glScalef (1.0f, -1.0f, 1.0f);
glTranslatef (0.0f, -1.0f, 0.0f);
glMatrixMode (GL_MODELVIEW);


Should I not do this and "flip" the coordinates on the fly?

Also one other thing, I see in your example code you "flip" vertex like me for ASE 3DS model, Vx=Vx,Vy=Vz, Vy=-Vy, but on the two sample MD5's I could fine
I've had to do Vx=Vy, Vy=Vz, Vz=Vx. Can you confirm that you do the first way, in which case it seems there's something else wrong with my code (even though it "looks" right, without textures :s ). Thanks

Share this post


Link to post
Share on other sites
I tried the texture matrix code you're posting and it's not what is causing the problems (anyway you should rather do loading texture coordinates and flipping them while loading - because these texture matrix operations must be done every frame (as long as you're using more than 1 model), but precomputing while loading not).

About that X Z-Y and Y Z X, you just mirror the model into some direction (so this won't cause issues). Although you have to use flipped axes at last (when you call glVertex, or as you generate vertex buffers - for joint position computing and for vertex position computing you have to use X Y Z - otherwise you can have sometimes X Y Z and sometimes X Z Y, etc. - and this will cause issues in vertex and its information transform).

So in final it really seems there is something more strange in the code...

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!