This repository was tested on Tensorrt 8.4.1-1 and CUDA-11.4
Create Engine Files
bash convertmodel.sh
Build Project
cd build && cmake .. && make -j
It builds 6 executibles
| Executible | Description |
|---|---|
| 1xDLAHybridInfer | 1 thread of DLA hybrid inference |
| 2xDLAHybridInfer | 2 threads of DLA hybrid inference |
| GPUOnlyInfer | 1 thread of GPU only inference |
| GPUwith1xDLAHybrid | 1 thread of GPU + DLA hybrid inference |
| GPUwith2xDLAHybrid | 2 threads of GPU + DLA hybrid inference |
| MaxThroughputFPSCount | 2 threads of GPU + DLA hybrid inference and 1 thread of GPU only |