Curvature-Driven Deformable Convolutional Networks for End-To-End Object Detection

Figure 4

Illustration of offset parts in (curvature-driven) deformable RoI pooling in R-FCN [11] and binsm(yellow) for an input RoI (red). Top: original RoI; the second row: DCNv1; the third row: DCNv2; bottom: our C-DCNets.