Post by thyme on Dec 4, 2004 6:38:33 GMT
Hi Everyone
Last year I found a good technique to compare the 3d proportions of my Petra avatar I was working on at the time to the photo of the real life model Petra Nemcova. I did this by taking a screen shot in paint shop pro of seamless's 3d window and pasting beside it a photo of the real life model. I had to scale my work to the photo so that I could draw horizontal lines from key points in the photo to the image of my 3d work. This proved to be extremely helpful in getting proportions correct. More recently I tried this with my current attempt in generating a realistic model from computer generated curves (using seamless build and generator nodes) and realised it is a lot easier to simply put the photo inside seamless's 3d window beside my 3d work instead of having to take a screen shot and align images to it in a separate program each time I want to compare my work to the photo. This is what I tried last night and I am really pleased with how much easier it is and the fact I could do all this without the need to add any new features to seamless
I simply added a new Seamless node as a sibling with a TextureEffect with a url field to the photo. I added a Part node and pasted a rectangle of triangles with a width and height ratio matching the width and height of the pixels for the photo. I then added the horizontal lines by adding extra Parts, each part containing 2 thin triangles to form the horizontal lines. Each line is easy to postilion like this using the Part's translation field. This morning I realised why not add another photo above and have vertical lines in this one so now both vertical and horizontal key points can be judged at anytime simultaneously.
Photos have limits, for example the head of the model Josie Maran shown here:
in the 3d window is probably tilted a little on the X axis and all photos will have some perspective which distorts proportions slightly. (I turn off perspective in seamless (go to settings)) but from this picture some very obvious truths are revealed. My nose is much shorter and narrower especially as it gets closer to the forehead. My eyes are centred further apart, are much bigger and more almond shaped. My head is much wider and my neck is much thinner. The mouth is a little harder to judge because in this photo Josie's mouth is slightly open. I think it could be fun and helpful to try and match a photo by matching the pose (such as opening the mouth a little in this case) but I also think you don't have to recreate the exact same image the important thing about this technique is it serves as a guide. Another thing I think exciting about putting the photo inside seamless is you can have the photo and the lines for it contained in a single seamless node like I do here and so it should be very easy to import different photo's of completely different models if you want to get an average or if you just want to compare the differences. Remember a single photo wont reveal all, different photos of the same model will look different depending on the mood, light, angle and the perspective of the photo and it is only a 2d representation of a 3d object so we would want different photos taken at different angles. We still have to fill in the blanks and guess some of it but I think this technique is a great help in aiding a better understanding of proportions.
If you have any questions relating to any of this or you have tried other techniques that give good results please reply
regards
thyme
Last year I found a good technique to compare the 3d proportions of my Petra avatar I was working on at the time to the photo of the real life model Petra Nemcova. I did this by taking a screen shot in paint shop pro of seamless's 3d window and pasting beside it a photo of the real life model. I had to scale my work to the photo so that I could draw horizontal lines from key points in the photo to the image of my 3d work. This proved to be extremely helpful in getting proportions correct. More recently I tried this with my current attempt in generating a realistic model from computer generated curves (using seamless build and generator nodes) and realised it is a lot easier to simply put the photo inside seamless's 3d window beside my 3d work instead of having to take a screen shot and align images to it in a separate program each time I want to compare my work to the photo. This is what I tried last night and I am really pleased with how much easier it is and the fact I could do all this without the need to add any new features to seamless
I simply added a new Seamless node as a sibling with a TextureEffect with a url field to the photo. I added a Part node and pasted a rectangle of triangles with a width and height ratio matching the width and height of the pixels for the photo. I then added the horizontal lines by adding extra Parts, each part containing 2 thin triangles to form the horizontal lines. Each line is easy to postilion like this using the Part's translation field. This morning I realised why not add another photo above and have vertical lines in this one so now both vertical and horizontal key points can be judged at anytime simultaneously.
Photos have limits, for example the head of the model Josie Maran shown here:
in the 3d window is probably tilted a little on the X axis and all photos will have some perspective which distorts proportions slightly. (I turn off perspective in seamless (go to settings)) but from this picture some very obvious truths are revealed. My nose is much shorter and narrower especially as it gets closer to the forehead. My eyes are centred further apart, are much bigger and more almond shaped. My head is much wider and my neck is much thinner. The mouth is a little harder to judge because in this photo Josie's mouth is slightly open. I think it could be fun and helpful to try and match a photo by matching the pose (such as opening the mouth a little in this case) but I also think you don't have to recreate the exact same image the important thing about this technique is it serves as a guide. Another thing I think exciting about putting the photo inside seamless is you can have the photo and the lines for it contained in a single seamless node like I do here and so it should be very easy to import different photo's of completely different models if you want to get an average or if you just want to compare the differences. Remember a single photo wont reveal all, different photos of the same model will look different depending on the mood, light, angle and the perspective of the photo and it is only a 2d representation of a 3d object so we would want different photos taken at different angles. We still have to fill in the blanks and guess some of it but I think this technique is a great help in aiding a better understanding of proportions.
If you have any questions relating to any of this or you have tried other techniques that give good results please reply
regards
thyme