| Channel: 3dor11 |
Your post is to long. Please make it short.
new posts available!
| 3DOR'11 is over. It was interesting to see all that work done by different groups. To shortly summarize: Heat Kernel Signatures as also Bag-of-Words approaches seems to be favourized in Shape Retrieval community! | |||
|
| the shrec'11 tracks summarize results of the benchmark. | |||
|
| last session is commited to SHREC Benchmarks. First talk in last session: SHREC'11 Track - Generic Shape Retrieval | |||
|
| next: Poster Session Fast Forward | |||
|
| next work is similar to the previous one, however in this case authors evaluate different feature point detectors. As ground truth they use "human-based" detector (people were asked to choose feature points). | |||
|
| interesting result of the previous work is, that mean curvature descriptors seams to work best, which is contradict our experience. I suppose this is because of the lack of non-rigid deformation experiments. | |||
|
| next one: Evaluation of 3D Interest Point Detection Techniques | |||
|
| the paper provides a comparison between different local shape descriptors. Data sets used include biological as well mechanical data. | |||
|
| next paper: Local shape descriptors, a survey and evaluation. | |||
|
| uniformly distributed interest points combined with local fourier spectrum descriptors are used to compute bag of features. afterwards this codebook is used to compare shapes | |||
|
| first talk after lunch: Bag of Words and Local Spectral Descriptor for 3D Partial Shape Retrievel | |||
|
| first talk after lunch: Bag of Words and Local Spectral Descriptor for 3D Partial Shape Retrievel | |||
|
| authors use heat kernels for feature detection (local maxima of autodiffusion function) and heat kernel signatures as descriptors. Training classifiers helps to classify feature points. application: autom. measurement of a human. | |||
|
| next talk: Heat diffusion approach for feature-based body scans annalysis. | |||
|
| the work combines graph matching with conformal factors as descriptors (local score of a region how much "power" you need to deform it to a sphere - so I suppose it measures uniformity of angles between normals). | |||
|
| next talk: ConTopo: Non-Rigid 3D Object Retrieval using Topological Information guided by Conformal Factors | |||
|
| a method for shape matching in biology, so even for non-isometric deformation. do this by mapping a surface to a grid plane and compute deformation of the parametrisation in the new domain, so "undo the stretching". | |||
|
| next talk: Refining shape correspondence for similar objects using strain | |||
|
| a hardware device using line strip method in infrared domain in order to retrieve faces. the system includes a software part to fill holes. system developed from scratch. system do eye tracking to decide when to take a snapshot for 3d reconstruction. | |||
|
| next talk Real-Time 3D Face Recognition using Line Projection and Mesh Sampling. | |||
|
| Keynote speaker today is Prof. Michael Bronstein. He gave an overview over the heat kernel descriptors and extensions for it they are working on,i.e.e incorporating color and volume. | |||
|