Linux distro and version: 18.04
GPU type Titan V
nvidia driver version: 4:10
CUDA version 10.0
CUDNN version 7.5
Python version [if using python] 3.6
Tensorflow version 1.14 nightly
TensorRT version 5.1.5
This layer is used by many object detection and semantic segmentation models and not having it really decreases the usability of tensorrt. Yes adding it as a plugin works but plugins should be for really uncommon / custom layers and not for common operations like upsampling / resizing.
Please add support for it in later releases or if there is a common workaround document it in a visible place or sample.