Deformable Part-based Fully Convolutional Network for Object Detection

Abstract

Existing region-based object detectors are limited to regions with fixed box\ngeometry to represent objects, even if those are highly non-rectangular. In\nthis paper we introduce DP-FCN, a deep model for object detection which\nexplicitly adapts to shapes of objects with deformable parts. Without\nadditional annotations, it learns to focus on discriminative elements and to\nalign them, and simultaneously brings more invariance for classification and\ngeometric information to refine localization. DP-FCN is composed of three main\nmodules: a Fully Convolutional Network to efficiently maintain spatial\nresolution, a deformable part-based RoI pooling layer to optimize positions of\nparts and build invariance, and a deformation-aware localization module\nexplicitly exploiting displacements of parts to improve accuracy of bounding\nbox regression. We experimentally validate our model and show significant\ngains. DP-FCN achieves state-of-the-art performances of 83.1% and 80.9% on\nPASCAL VOC 2007 and 2012 with VOC data only.\n