The research within several of our projects incorporates creation of fully operating prototypes of index and search structures and applications. The software is typically written in Java, individual modules are based on the same platform and can cooperate. Majority of the SW packages are accessible at Bitbucket DISA Lab.
- MESSIF (Metric Similarity Search Implementation Framework) – framework designed to support the process of building prototypes of metric-based indexeding techniques
- M-Tree – implementation of famous dynamic indexing structure for metric-based searching
- D-Index – implementation of a hashing schema for similarity queries including similarity joins
- M-Index – a precise and approximate metric-based index based on principles of hashing data to a single-dimensional domain
- SMF – Subsequence Matching Framework (SMF) is a versatile framework for prototyping, building and testing applications employing subsequence matching approaches.
- PPP-Codes – a new fast approximate metric-based index
License: These software modules are free software: you can redistribute them and/or modify them under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
The following three pieces of software have been developed under the project Efficient Searching in Large Biometric Data (No. VG20122015073) within the security-research programme (BV II/2-VS) funded by Ministry of the Interior of the Czech Republic.
- FaceMatch – a face retrieval technology integrating multiple face detection and matching methods
- MotionMatch – a person identification technology based on the way people walk
- MMPI (Multi-modal Person Identification) – a multi-modal person identification technology based on similarity of faces and the way people walk
In several of our projects, we study the processing of multimedia data in the context of internet. To illustrate the effectiveness and efficiency of our methods, we offer some of the similarity search functionalities in the form of plugins for Mozilla Firefox. With no need to compile source codes and understand our technologies, you can easily try similarity-based image search and annotation using the following plugins:
Data collections that can be used for evaluation purposes are an important asset for the scientific community. In the course of our research of image retrieval techniques, we have created a large collection of images and image descriptors, that we offer to other research groups for evaluation and benchmarking:
In the course of our research in the area of image annotation, we have also created a new ontology for image content description. The Visual Concept Ontology organizes the most common concepts that appear in general images and links them to relevant objects in the WordNet lexical database.