Hi,
I am implementing roi pooling layer (see faster rcnn) in tensorRT.
the dim returned by getOutputDimensions is Dims(1000, 256, 7, 7).
something like
Dims getOutputDimensions(int index, const Dims* inputs, int nbInputDims) override
{
LOG(INFO) << "roi pool getOutputDimensions";
return Dims4(1000, 256, 7, 7);
}
and there are a fully connected layer which is followed by a batch norm layer.
layer {
name: "roi_pooling"
type: "IPlugin"
bottom: "p2"
bottom: "p3"
bottom: "p4"
bottom: "p5"
bottom: "all_rois"
bottom: "all_scores"
top: "all_roi_pooling"
}
layer {
name: "new_fc6"
type: "InnerProduct"
bottom: "all_roi_pooling"
top: "new_fc6"
param {
name: "fc6_w"
lr_mult: 1
}
param {
name: "fc6_b"
lr_mult: 2
}
inner_product_param {
num_output: 1024
}
}
layer {
name: "new_fc6/bn"
type: "BatchNorm"
bottom: "new_fc6"
top: "new_fc6"
param {
lr_mult: 0
}
param {
lr_mult: 0
}
param {
lr_mult: 0
}
batch_norm_param {
use_global_stats: true
}
}
However, when parsing the network
tensorRT throws the following errors
ERROR: Parameter check failed at: ../builder/Network.cpp::addScale::120, condition: hasChannelDimension == true
error parsing layer type BatchNorm index 430
any idea?