Over the last days I worked on the ModelConverterX LOD Creator tool. My first aim was to add another algorithm to calculate the LODs. And besides that I also improved the user interface a bit.
On the right you see a screenshot of the new user interface and since the manual is not yet updated, I will give a brief overview of the changes here.
The selection box at the top right allows you to choose between the different algorithms that are available. At the moment you can choose either vertex clustering or quadratic based error. The last one is the algorithm that was used in the previous version of ModelConverterX and it has not changed in this new update. I hope to improve that algorithm more later on. Below the algorithm selection box you see the different parameters that you can set for each of them.
The remaining changes in the user interface are mainly the buttons on the bottom left. They allow you to remove a LOD, rename a LOD or add an empty LOD. New is the import LOD button which allows you to load another file and that object will then be used for the LOD number indicated next to the button. This should allow people to make for example LODs with SketchUp. By saving them all to different files, you can combine them in one MDL using this tool.
Then to the new algorithm. What vertex clustering basically does is collapse all vertices that are within the same cell of a 3D grid. The resulting vertex position depends on the weight of the original vertices. In the implementation into ModelConverterX you can give vertices a weight based on:
- The area of the polygons that the vertex is a part of. This means that vertices that are part of a big polygon will have more influence on the new shape. In the properties you can set the influence of the area at a value between 0 and 1. Where 0 means no additional weight and 1 means the maximum additional weight.
- The angle between the polygons that the vertex is a part of. The idea behind this weight is that polygons on the outside of the object are more likely to have sharp edges. By given them more weight the overall silhouette will be preserved better. This parameter also has a range between 0 and 1.
- Ground clamping. This gives an additional weight to vertices that are on or below the ground. This is to ensure that you building does not float in the lower LODs.
The two remaining parameters of the algorithm are grid size and floating cells. The first one is quite obvious and determines how big the cells are, within which the vertices are collapsed. The second allows you to choose between uniform fixed cells and floating cells. According to the literature the last option should give better results in general, although I have already seen that it does not apply to every object.
So I hope you enjoy these new LOD functionalities and if you want to see them in action have a look at this demonstration video.