Skip to content
No description, website, or topics provided.
MATLAB M
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
dependencies
meshes
.gitattributes
QSlim.exe
README.md
autocrop.m
check_face_vertex.m
cnn_shape_view.m
cnn_view_output.m
compute_mesh_weight.m
deploysaliency.m
get_augmentation_matrix.m
getoptions.m
icosahedron2sphere.m
loadMesh.m
mark_visible_vertices.m
mrf4.m
perform_mesh_simplification.m
perform_mesh_smoothing.m
plotMesh.m
ray_trianle_intersect.m
read_smf.m
readoff.m
render_views.m
saliencytrain.m
same_side.m
setup.m
triangulation2adjacency.m
vring.mat
write_smf.m

README.md

The source codes are based on matconvnet.

  1. Creat a new directory \data\models

  2. Download the pretrained net from

https://www.dropbox.com/s/15glsp57wp7qgj3/net-deployed.mat?dl=0

  1. Save the downloaded net file 'net-deployed.mat' in \data\models

For a demo, simply implement the following line in MATLAB command window

s1=deploysaliency('.\meshes\human.off');

You can also input a scene:

s2=deploysaliency('.\meshes\conferenceroom.off');

To calculate the distinction of your own 3D mesh/scene, if you can find a mesh of the same object class in the 'meshes' directory, make sure that the orientation of your mesh is roughly the same as that of the corresponding one in the 'meshes' directory. Otherwise just make it up oriented.

Please cite our paper if you use the codes:

Ran Song, Yonghuai Liu, Paul L. Rosin. Distinction of 3D Objects and Scenes via Classification Network and Markov Random Field. IEEE Transactions on Visualization and Computer Graphics, 15 pages, 2018

You can’t perform that action at this time.