. "N\u00E1vrh klasifik\u00E1toru dopravn\u00EDch zna\u010Dek zalo\u017Een\u00E9ho na podobnosti zkouman\u00E9ho objektu ke t\u0159\u00EDd\u011B zna\u010Dek reprezentovan\u00E9 v\u017Edy typickou zna\u010Dkou (prototype-based rule). Navr\u017Een algoritmus zalo\u017Een\u00FD na tr\u00E9nov\u00E1n\u00ED podle mno\u017Einy prototyp\u016F (trainable similarity). Experimenty na n\u011Bkolika datov\u00FDch souborech ilustruj\u00ED vy\u0161\u0161\u00ED \u00FA\u010Dinnost klasifik\u00E1toru v porovn\u00E1n\u00ED s klasifik\u00E1tory a\u017E dosud pou\u017E\u00EDvan\u00FDmi pro klasifikaci zna\u010Dek. Byly testov\u00E1ny i dal\u0161\u00ED vlastnosti navr\u017Een\u00E9ho klasifik\u00E1toru jako robustnost a \u010Dasov\u00E1 n\u00E1ro\u010Dnost."@cs . . "Building Road-Sign Classifiers Using a Trainable Similarity Measure" . "A frequently used strategy for road sign classification is based on the normalized cross-correlation similarity to class prototypes followed by the nearest neighbor classifier. Because of the global nature of the cross-correlation similarity, this method suffers from presence of uninformative pixels (caused e.g. by occlusions), and is computationally demanding. In this paper, a novel concept of a trainable similarity measure is introduced which alleviates these shortcomings. The similarity is based on individual matches in a set of local image regions. The set of regions, relevant for a particular similarity assessment, is refined by the training process. It is illustrated on a set of experiments with road sign classification problems that the trainable similarity yields high-performance data representations and classifiers. Apart from a multi-class classification accuracy, also non-sign rejection capability, and computational demands in execution are discussed. It appears that the trainable simi..."@en . . . . "US - Spojen\u00E9 st\u00E1ty americk\u00E9" . "P(2C06019), P(IAA2075302), R, Z(AV0Z10750506)" . "IEEE Transactions on Intelligent Transportation Systems" . "3" . "RIV/67985556:_____/06:00041079" . "Klasifikace dopravn\u00EDch zna\u010Dek zalo\u017Een\u00E1 na m\u00ED\u0159e podobnosti zkouman\u00E9ho objektu k t\u0159\u00EDd\u011B reprezentovan\u00E9 typickou zna\u010Dkou"@cs . "13"^^ . "309;321" . . "7" . . . . "467503" . . "1524-9050" . "RIV/67985556:_____/06:00041079!RIV07-AV0-67985556" . "classifier system design; road-sign classification; similarity data representation"@en . . "Klasifikace dopravn\u00EDch zna\u010Dek zalo\u017Een\u00E1 na m\u00ED\u0159e podobnosti zkouman\u00E9ho objektu k t\u0159\u00EDd\u011B reprezentovan\u00E9 typickou zna\u010Dkou"@cs . . . "Pacl\u00EDk, P." . "Novovi\u010Dov\u00E1, Jana" . "3"^^ . . . "Building Road-Sign Classifiers Using a Trainable Similarity Measure" . . "Duin, R. P. W." . "1"^^ . "[3DEDEC319B33]" . "Building Road-Sign Classifiers Using a Trainable Similarity Measure"@en . . "A frequently used strategy for road sign classification is based on the normalized cross-correlation similarity to class prototypes followed by the nearest neighbor classifier. Because of the global nature of the cross-correlation similarity, this method suffers from presence of uninformative pixels (caused e.g. by occlusions), and is computationally demanding. In this paper, a novel concept of a trainable similarity measure is introduced which alleviates these shortcomings. The similarity is based on individual matches in a set of local image regions. The set of regions, relevant for a particular similarity assessment, is refined by the training process. It is illustrated on a set of experiments with road sign classification problems that the trainable similarity yields high-performance data representations and classifiers. Apart from a multi-class classification accuracy, also non-sign rejection capability, and computational demands in execution are discussed. It appears that the trainable simi..." . . . "Building Road-Sign Classifiers Using a Trainable Similarity Measure"@en .