Rong Tianze ; Zhu Yanjia ; Cai Hongxiang ; Xiong Yichao

Abstract

This work is a solution to densely packed scenes dataset SKU-110k. Our workis modified from Cascade R-CNN. To solve the problem, we proposed a random cropstrategy to ensure both the sampling rate and input scale is relativelysufficient as a contrast to the regular random crop. And we adopted some oftrick and optimized the hyper-parameters. To grasp the essential feature of thedensely packed scenes, we analysis the stages of a detector and investigate thebottleneck which limits the performance. As a result, our method obtains 58.7mAP on test set of SKU-110k.