In this work, we present “Qatris IManager”, a powerful software to search, edit and classify images, belonging to SICUBO S.L.; a spin-off from University of Extremadura and located in Cáceres (Spain). “Qatris IManager” manages and classifies images using four kinds of features: color, texture, shape and user description.
Qatris Imanager is a solution created to home users. It allows to manage image collections with categories customized by the user, integrating automatic loading and search by content tools.
Qatris IManager search and download images from websites with customized bots, to do this it extracts images features and classifies them automatically incustomized categories. Classification can be performed with supervised (modified bayesian logistic regression) and non-supervised (clusters) statistical methods. Using a prior classification of images made by the user, the system applies this information to recognize and classify new images.
- Automatic Images Classification in customized categories.
- Automatic Images Download from Websites with customized bots.
- Image Matching by similarity measures.
- Interactive Learning with relevance feedback.
Technology in Qatris IManager
We implement algorithms for image features extraction (color, texture, shape and description features), to use the most representative ones for each application. These are algorithms based on color, textures and shape to be applied on full images or regions of interest (ROIS) obtained from grid division or by stressed contour detection.
We produce hybrid indexing systems, mixing classic techniques (as relation alone) with other more innovative as multidimensional indexing of features vectors. With this technology we achieve to store embedded digital documents in databases, being accessible in different ways: by key, metric or meta-data,...
We develop statistical analysis tools to process features vectors extracted from images (or other documents), search similar images to the one you selected by color, texture and shape features and classify them with supervised and non supervised patterns recognition methods.
Below is a series of screenshots of the application Qatris iManager.
With Qatris Imanager you can easily classify or search images by visual content. For classification you can use two different methods: supervised and non supervised. Non supervised method only needs an initial sample of unclassified images to classify in a number of classes decided by the user. This method is some times less intuitive and can classify by features that human eye usually skips. Supervised method uses revised by the user classification; it means that the user must give a pre classified sample of images for each class he wants to create.
If you already have the images sorted into folders, you can create the classifier based on these folders.
For classification we use modified bayesian logistic regression achieving a high accuracy. The system classifies images from linked folders and put them in to classes or ask the user to supervise them.
With an easy and intuitive interface you can supervise classification results and correct them if it is necessary. If the classifier was configured to add new images classified to the initial sample, the system will update the statistical weights and it will learn from the new information to adapt the results to the user´s point of view.
This method has a feed back learning method which allows the system to know what the user really wants to classify by.
Qatris also performs searches by content trying to identify which images are similar to the one given by the user learning his point of view.
For this customized search we use the pairwise comparison method and relevance feed back learning algorithm. When the results of the search are shown, the user decides which images are similar to the one he has selected and which are not. We take this information and make the system understands the opinion of the user to fit as much as possible in the next search.
Not only you can search similar images to the one you chose, but also Qatris Imanager uses tools to create aided visual descriptions dividing the image in a grid to make an intuitive search.
With this kind of search you will select the color, texture and shape features you are looking for in each zone of the grid.
We currently have available the next version of iManager Qatris for download.
Qatris IManager version 2.0. Available in both 32 and 64 bit as well as in Spanish and English. If you have questions about which architecture to choose download and use the 32 bits version.
|Qatris IManager 32 bits||English||
|Qatris IManager 64 bits||English||
|Qatris IManager 32 bits||Spanish||
|Qatris IManager 64 bits||Spanish||
Qatris IManager version 1.0.