This is TensorRT version of efficientnetB3, support trained tensorflow 2.x weights
TensorRT 7.2.3
cmake>3.2
tensorflow>2.0
keras efficientnet 1.1
cuda 11.0
1.Install all required package
2.Run efficientnet.py (you will get a pretrained model and generate weights to wts file (NCHW) format)
3.config paths in cmakelists.txt
1.gen a fir using
mkdir build &&cd build
and run
cmake ..&&make
to gen exec.
2. generate engine, using
./effficientnet -s
command.
3. test random generated imgs, using
./efficientnet -d