Department of Computer Science Technische Universitaet Muenchen
Home People Teaching Research Publications Info Sitemap
About | Teaching | Research

Automatic shape model inference from google sketchup

For the automatic recognition of objects in a human populated environment shape models are an interesting starting point. For most of the daily used objects in our environment there exist rough shape models in the internet. We need to extract for a given word, where we already some the semantic information about an approximated shape model with a small number of degrees of freedom. This can be done using some kind of fusion of 3D Models. To do such a fusion techniques like PCA or other kind of component analysis should be used in order to get highest use of the various models that can be found in google sketchup. Finding such models in google sketchup is also an interesting task, since a lot of noise in form of semantical mismatches must be filtered out. We already automatically extract shape models from the internet given a name of it, we can generate new intermediate model given a set of results. We are able to search for those models in images. Have a look to the clustering result which sorts out uninteresting models. For the final recognition there is still work to do: the large variety of models results often with false positive matches. To sort them out a list of techniques can be applied.


Computer Science IX | Technische Universität München