Is it possible to add custom layers to tensorRT? e.x. the “proposal_layer” and “ROI_pooling_layer” of the Faster-Rcnn
Best