Network: The network must contain one ROI max pooling layer

I built a faster-RCNN network using AlexNet for feature extraction. Generally this detector network use one ROI max pooling layer, but I want to use two ROI max pooling layers. As, in my case, one default ROI max pooling after relu5 and one additional ROI max pooling after relu2. Then output of these two ROI pooling layers are merged element wise. When i try to tarin this network with "trainFasterRCNNObjectDetector()" function, it throws an error as mentioned in title. How I can use two ROI maxpooling layers. The code file is also attached for better understanding.

Thanks for helping out.