In this paper, a new scalable configuration is introduced for arranging Feature Pyramid Networks (FPNs) known as Vertical FPNs or VFPNs. Our work is an alternative approach to the one commonly employed for computer vision tasks. The FPNs are parallelized for improved learning of complex features directly from backbone. Our hypothesis is validated by parallelizing EfficientDets FPNs. This method of scaling is an effective method to boost the performance and improve the mAP.