Graphics Reference
In-Depth Information
1.
Automatic Location of Landmarks used in Manual Anthropometry (
http://www.andrea
giachetti.it/shrec14/
);
2.
Shape Retrieval of Non-Rigid 3D Human Models (
http://www.cs.cf.ac.uk/shaper
etrieval/shrec14/index.html
);
3.
Retrieval and Classification on Textured 3D Models (
http://www.ge.imati.cnr.it/sh
rec14
);
4.
Extended Large Scale Sketch-Based 3D Shape Retrieval (
http://www.itl.nist.gov/i
ad/vug/sharp/contest/2014/SBR/
);
5.
Large Scale Comprehensive 3D Shape Retrieval (
http://www.itl.nist.gov/iad/vug
/sharp/contest/2014/Generic3D/
).
Segmentation
e mesh segmentation benchmark
http://segeval.cs.princeton.edu/
provides
380
meshes across
19
object categories for quantitative analysis of how people decom-
pose objects into parts and for comparison of automatic mesh segmentation algorithms. To build
the benchmark, eighty people were recruited to manually segment surface meshes into functional
parts, yielding an average of
11
human-generated segmentations for each mesh. is data set pro-
vides a sampled distribution over “how humans decompose each mesh into functional parts,” and
this knowledge is treated as a probabilistic “ground truth.” In addition, the dataset is coupled with
software for evaluation, analysis, and viewing of mesh segmentations. e code is written in C++
and free to use. Also Python scripts to automatically run evaluation and analysis experiments are
available to plot the results (using Matlab), to create reports, as well as to generate colored images
of mesh segmentations. e idea is that this benchmark can be used to study and compare new
segmentation algorithms, provided that the results of seven methods are given together with the
human-generated mesh segmentations. e survey [
45
] demonstrates that there is not a single
algorithm that performs well on all shape classes.
Another shape segmentation benchmark is the 3D Segmentation Benchmark [
15
] avail-
able at
http://www-rech.telecom-lille1.eu:8080/3dsegbenchmark/
. Similarly to [
45
],
this benchmark provides twenty-eight 3D-models grouped in five classes, namely animal, furni-
ture, hand, human and bust. Each 3D-model is associated with some manual segmentations done
by volunteers. A small number of varied models with respect to a set of properties is selected. All
the selected models are manifold, connected, and do not have intersecting faces. Hence they are
supported as an input by any segmentation algorithm. In order to collect precise manual segmen-
tations, the volunteers have been assisted when traced the vertex-boundaries through the different
models even if no condition was imposed on the manner with which they have segmented them.
To evaluate a new segmentation algorithm it is necessary to create a user account, upload
the files which are the results of the segmentation algorithm and the evaluation result are displayed
online, in a temporary page. In addition, the results of eight segmentation algorithms are available
Search WWH ::
Custom Search