Figure 5

The architecture of Cascade R-CNN w/ACASEM. P represents object proposals from the RPN, \(N_{j}\) is the network body, \(L_{j}\) are class labels predictions and \(B_{j}\) are the bounding box predictions.

The architecture of Cascade R-CNN w/ACASEM. P represents object proposals from the RPN, \(N_{j}\) is the network body, \(L_{j}\) are class labels predictions and \(B_{j}\) are the bounding box predictions.