Custom processor on 2-line license plate recognition

Accelerated Computing Intelligent Video Analytics DeepStream SDK
17

I read this post and saw that the post owner use tee to split into 2 branches:

https://forums.developer.nvidia.com/t/is-the-lprnet-trainable-for-license-plates-with-two-separate-lines-instead-of-single-one-as-of-us/169648/13?u=hungtv

PGIE --> SGIE_1 (LPDnet)|--> SGIE_2 (LPRnet_1)
                        |--> SGIE_3 (LPRnet_2)

I tried that but I don’t know how to merge the SGIE_2 and SGIE_3 together, is that an element I can use to merge 2 branches? Or I need to write some C++ code to work on that?