UDLF Versions Save

• Bug fixes

New features:

• Add automatic input file format detection.

Fixes:

• Fix CONTEXTRR method;
• Fix RDPAC for matrix input;
• Fix BFSTREE for matrix input.

• The user can compute multiple recalls at once;
• Dataset size is now long int, which avoids memory corruption (wrong indexing) for larger datasets.

The RDPAC (Rank-based Diffusion Process with Assured Convergence) re-ranking algorithm is now available in UDLF. Enjoy!

Rank-based Diffusion Process with Assured Convergence (RDPAC) is presented in the paper: PEDRONETTE, D. C. G.; VALEM, L. P.; LATECKI, L. J. . "Efficient Rank-based Diffusion Process with Assured Convergence." Journal of Imaging (ISSN 2313-433X). https://doi.org/10.3390/jimaging7030049

The BFSTree re-ranking algorithm is now available in UDLF. Enjoy! DOI: https://doi.org/10.1016/j.patcog.2020.107666

The Log-based Hypergraph of Ranking References (LHRR) is now available in UDLF. Enjoy! DOI: http://doi.org/10.1109/TIP.2019.2920526

Bug fix:

• If the user hasn't declared a parameter in the configuration file, the program could crash when trying to set it to the default value. This bug could be reproduced when the binary was compiled with earlier versions of the G++ compiler (4.8.4, for example).

The internal configuration files are now embedded into the binary at compile time. As an advantage, the binary can be executed from any directory.

A Stable Release of the Unsupervised Distance Learning Framework