I am implementing roi pooling in tensorRT.
after roi pooling layer, the output dimension is NCHW.
for example, with 1000 proposals, 256 channels and both pooling height and pooling width are 7.
the output dimension would be (1000, 256, 7, 7).
the following layer is FC layer.
However, If I write FC layer as plugin layer, the batchSize in enqueue is still 1, not 1000.
in this circumstance, only the first roi would be computed.
I am not sure what is the correct way to do that.