TensorRT- unable to parse layers in SSD_VGG network

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi,

I am trying to run SSD_VGG model trained on ILSVRC dataset. ( SSD300 model - link at bottom of page GitHub - weiliu89/caffe at ssd ). TensorRT is not able to parse the normalization layer (conv4_3_norm ) and the parameters present in this layer ( norm_param ). Is there a solution for this issue ? ( renaming the layer/parameters etc.) TensorRT version used is 2.1

2

Same problem even with TensorRT 3.0.

3

Yeah, using 3.0.4 and also getting the error:

Message type "ditcaffe.LayerParameter" has no field named "norm_param".

I guess we would have to write a custom layer for it ?

4

I found a github discussion: Mobilenetssd · Issue #5 · chenzhi1992/TensorRT-SSD · GitHub
The issue is discussed near the end of the discussion.
Hope it helps