M. Godi, M. Carletti, M. Aghaei, F. Giuliari, M. Cristani
Project @ author's page
NOTE
The project consists of two parts. Given a set of images belonging to the same class/category, the former part generates a crisp saliency mask for each image in the set. The second part computes a set of visual summaries starting from the crisp masks.
This is the SECOND part of the project.
You can find HERE the first part of the project concerning the computation of the crisp masks.
To generate crisp saliency maps (first part) you need to follow the instructions here.
To generate a set of visual summaries (second part) for a specified class you need to:
- Install the MATLAB software (tested on Matlab2017b).
- Follow the installation instructions for Proposal Flow with SelectiveSearch option as proposal method
- Set the path to ProposalFlow installation folder in 'set_proposal_flow_folder.m'
- Set 'in_base_folder' and 'out_base_folder' as, respectively, input and ouput folders in 'compute_clusters.m'; 'in_base_folder' should contain a folder for each class (each one produced by running the first part of the project) for which the summaries have to be computed.
Follow the instructions here.
Run 'compute_clusters.m' to generate visual summaries. The resulting images are going to be separated into folders by class in 'out_base_folder', together with corresponding mat files that can be used to speed up future computations (to recompute summaries after changing parameters, remove the .mat files)