search

dusty-nv / jetson-inference

Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.
https://developer.nvidia.com/embedded/twodaystoademo
MIT License
7.87k stars 2.98k forks source link

I want to add my customized layer #24

Closed SeougnSeon closed 7 years ago

SeougnSeon commented 7 years ago

I want to add my customized layer in Jetson-Inference demo.

Currently, GIE does not support customized layers as well as latest caffe versions.

I want to add my caffe distribute(libcaffe.so) in this camera demo by removing GIE.

Is it possible?

suhyung commented 7 years ago

I also strongly hope this feature

e2sh83 commented 7 years ago

Can anybody answer for this request? It would be great if applying customized layer is possible.

dusty-nv commented 7 years ago

Hi there, custom layers are a feature of TensorRT2, the next version.

For now, to add support for custom layers, please use Caffe or your own code. The Tegra Multimedia API which JetPack installs, includes a sample of using TensorRT and cuDNN in tandem to implement a 'custom' function.

For example, the detectNet and segNet example networks that are included in this repo, I removed superfluous layers to get it working. Typically these are at the end of the pipeline and are linear operators (like cropping, ect.) as opposed to having non-zero learning rate.

SeougnSeon commented 7 years ago

Thank you for your kind explanation.