Sparse weakly supervised models for object localization in road environment (CROSBI ID 260101)
Prilog u časopisu | izvorni znanstveni rad | međunarodna recenzija
Podaci o odgovornosti
Zadrija, Valentina ; Krapac, Josip ; Šegvić, Siniša ; Verbeek, Jakob
engleski
Sparse weakly supervised models for object localization in road environment
We propose a novel weakly supervised localization method based on Fisher-embedding of low-level features (CNN, SIFT), and model sparsity at the component level. Fisher-embedding provides an interesting alternative to raw low-level features, since it allows fast and accurate scoring of image subwindows with a model trained on entire images. Model sparsity reduces overfitting and enables fast evaluation. We also propose two new techniques for improving performance when our method is combined with nonlinear normalizations of the aggregated Fisher representation of the image. These techniques are (i) intra-component metric normalization and (ii) first-order approximation to the score of a normalized image representation. We evaluate our weakly supervised localization method on real traffic scenes acquired from driver’s perspective. The method dramatically improves the localization AP over the dense non-normalized Fisher vector baseline (16 percentage points for zebra crossings, 21 percentage points for traffic signs) and leads to a huge gain in execution speed (91x for zebra crossings, 74x for traffic signs).
Object localization Weak supervision Fisher vectors Sparse models Convolutional features Geographic information system (GIS) OpenStreetMap
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o izdanju
176–177
2018.
9-21
objavljeno
1077-3142
1090-235X
10.1016/j.cviu.2018.10.004