Cordts, Marius ; Rehfeld, Timo ; Enzweiler, Markus ; Franke, Uwe ; Roth, Stefan (2017)
Tree-Structured Models for Efficient Multi-Cue Scene Labeling.
In: IEEE Transactions on Pattern Analysis and Machine Intelligence, 39 (7)
doi: 10.1109/TPAMI.2016.2592911
Artikel, Bibliographie
Kurzbeschreibung (Abstract)
We propose a novel approach to semantic scene labeling in urban scenarios, which aims to combine excellent recognition performance with highest levels of computational efficiency. To that end, we exploit efficient tree-structured models on two levels: pixels and superpixels. At the pixel level, we propose to unify pixel labeling and the extraction of semantic texton features within a single architecture, so-called encode-and-classify trees. At the superpixel level, we put forward a multi-cue segmentation tree that groups superpixels at multiple granularities. Through learning, the segmentation tree effectively exploits and aggregates a wide range of complementary information present in the data. A tree-structured CRF is then used to jointly infer the labels of all regions across the tree. Finally, we introduce a novel object-centric evaluation method that specifically addresses the urban setting with its strongly varying object scales. Our experiments demonstrate competitive labeling performance compared to the state of the art, while achieving near real-time frame rates of up to 20 fps.
| Typ des Eintrags: | Artikel |
|---|---|
| Erschienen: | 2017 |
| Autor(en): | Cordts, Marius ; Rehfeld, Timo ; Enzweiler, Markus ; Franke, Uwe ; Roth, Stefan |
| Art des Eintrags: | Bibliographie |
| Titel: | Tree-Structured Models for Efficient Multi-Cue Scene Labeling |
| Sprache: | Englisch |
| Publikationsjahr: | 2017 |
| Titel der Zeitschrift, Zeitung oder Schriftenreihe: | IEEE Transactions on Pattern Analysis and Machine Intelligence |
| Jahrgang/Volume einer Zeitschrift: | 39 |
| (Heft-)Nummer: | 7 |
| DOI: | 10.1109/TPAMI.2016.2592911 |
| URL / URN: | https://doi.org/10.1109/TPAMI.2016.2592911 |
| Kurzbeschreibung (Abstract): | We propose a novel approach to semantic scene labeling in urban scenarios, which aims to combine excellent recognition performance with highest levels of computational efficiency. To that end, we exploit efficient tree-structured models on two levels: pixels and superpixels. At the pixel level, we propose to unify pixel labeling and the extraction of semantic texton features within a single architecture, so-called encode-and-classify trees. At the superpixel level, we put forward a multi-cue segmentation tree that groups superpixels at multiple granularities. Through learning, the segmentation tree effectively exploits and aggregates a wide range of complementary information present in the data. A tree-structured CRF is then used to jointly infer the labels of all regions across the tree. Finally, we introduce a novel object-centric evaluation method that specifically addresses the urban setting with its strongly varying object scales. Our experiments demonstrate competitive labeling performance compared to the state of the art, while achieving near real-time frame rates of up to 20 fps. |
| Freie Schlagworte: | Automotive industries, Labeling, Feature extraction |
| Fachbereich(e)/-gebiet(e): | 20 Fachbereich Informatik 20 Fachbereich Informatik > Graphisch-Interaktive Systeme |
| Hinterlegungsdatum: | 05 Mai 2020 14:28 |
| Letzte Änderung: | 05 Mai 2020 14:28 |
| PPN: | |
| Export: | |
| Suche nach Titel in: | TUfind oder in Google |
 |
Frage zum Eintrag |
Optionen (nur für Redakteure)
 |
Redaktionelle Details anzeigen |
Drucken
Impressum