. "GAP103/12/2310" . . . "http://www.isvav.cz/projectDetail.do?rowId=GAP103/12/2310"^^ . . "We propose to use recent methods in image retrieval and clustering to obtain large amount (orders of magnitude more than were used so far) of high quality training image data for a number of computer vision tasks. Methods of machine learning that deal with large amounts of data in the training phase and are reasonably fast in the classication phase will be investigated. The machine learning methods will be used to learn classier of key-point quality and to learn similarity measures and clusterings in high dimensional descriptor spaces. To achieve the goal, further improvements in image retrieval will be pursued. This includes novel methods for automatic query expansion and complex repeated pattern analysis."@en . . . . "2014-12-31+01:00"^^ . "11"^^ . "11"^^ . . "2012-01-01+01:00"^^ . "Large Scale Image and Object Search as a Teacher"@en . . "Project's contributions are unquestionable in the area of image retrieval, which is documented by publications on outstanding conferences and journals (CVPR, ECCV, IJCV and TPAMI). During the project, J. Pritts graduated and has started PhD studies, both supervised by PI. Quality of results is supported by 25 citations in WoS database."@en . . . . "0"^^ . . . "Navrhujeme pou\u017E\u00EDt\u00ED aktu\u00E1ln\u00EDch metod vyhled\u00E1v\u00E1n\u00ED obraz\u016F a jejich shlukov\u00E1n\u00ED k z\u00EDsk\u00E1n\u00ED velk\u00E9ho mno\u017Estv\u00ED (v \u0159\u00E1dech v\u00EDce ne\u017E bylo pou\u017Eito dosud) vysoce kvalitn\u00EDch (tj. s mal\u00FD po\u010Dten chyb) trenovac\u00EDch dat pro r\u016Fzne probl\u00E9my po\u010D\u00EDta\u010Dov\u00E9ho vid\u011Bn\u00ED. Budou zkoum\u00E1ny metody strojov\u00E9ho u\u010Den\u00ED kter\u00E9 se dok\u00E1\u017E\u00ED vypo\u0159\u00E1dat s velk\u00FDm mno\u017Estv\u00EDm tr\u00E9novac\u00EDch dat a z\u00E1rove\u0148 jsou rozumn\u011B rych\u00E9 p\u0159i klasifikaci. Metody strojov\u00E9ho u\u010Den\u00ED budou aplikov\u00E1ny k u\u010Den\u00ED klasifik\u00E1toru kvality v\u00FDzna\u010Dn\u00FDch bod\u016F a k u\u010Den\u00ED podobnosti a shlukov\u00E1n\u00ED ve vysokodimenzion\u00E1ln\u00EDch prostorech descriptor\u016F. K dosa\u017Een\u00ED tohoto c\u00EDle budeme d\u00E1le vylep\u0161ovat metody vyhled\u00E1v\u00E1n\u00ED obraz\u016F. To zahrnuje nov\u00E9 metody pro automatickou expanzi dotazu, a studov\u00E1n\u00ED komplexn\u00EDch opakovan\u00FDch vzor\u016F." . "2014-04-18+02:00"^^ . "1"^^ . "image and object search machine learning key-point detection"@en . "2015-05-22+02:00"^^ . "Vyhled\u00E1v\u00E1n\u00ED obraz\u016F a objektu ve velk\u00FDch datab\u00E1z\u00EDch jako u\u010Ditel" . . "P\u0159\u00EDnos projektu je nezpochybniteln\u00FD v oblasti vyhled\u00E1v\u00E1n\u00ED obr\u00E1zk\u016F, co\u017E je dolo\u017Eeno v\u00FDsledky na vynikaj\u00EDc\u00EDch f\u00F3rech (CVPR, ECCV a \u010Dasopisy IJCV a TPAMI). B\u011Bhem \u0159e\u0161en\u00ED projektu J. Pritts graduoval v mgr. programu a zah\u00E1jil PhD studium, oboje pod veden\u00EDm \u0159e\u0161itele. Kvalita v\u00FDsledk\u016F projektu je podpo\u0159ena ji\u017E 25 citacemi ve WoS."@cs . . . . "0"^^ .