V-Sense / salnet360

Repository for implementation of SalNet360 in Caffe
MIT License
18 stars 3 forks source link
360-photo caffe deep-learning saliency saliency-map

salnet360

Repository for implementation of salnet360 in Caffe.

salnet360 creates a visual attention map for omni-directional/360 images.

For more details on the model refer to "SalNet360: Saliency maps for omni-directional images with CNN"

Model and Parameter Files

The model and parameters files in models directory.

vsenseSalNet360Deploy.prototxt is the file that defines the network model and vsenseSalNet360Final_iter_12000.caffemodel has the optimized parameters we found using the training method mentioned in the paper.

For creating a saliency map for omni-directional image use the following command

createSalMap360.sh \ \

Dependencies