Accelerate Pelee with TensorRT Pelee: A Real-Time Object Detection System on Mobile Devices (NeurIPS 2018)

TensorRT-Pelee can run over 70FPS(11ms) on Jetson TX2(FP32)

Performance(FP32)

1. NVIDIA Jetson TX2: 72 FPS (13.2~11 ms)
   
2. Titan V: 200 FPS (5 ms)
   

Requirements:

1. TensorRT 4.x (Jetpack 3.3)
   
2. CUDA 9.0
   
3. cudnn 7.
   

Run:

cmake .
make
./build/bin/pelee

Reference:

https://github.com/Ghustwb/MobileNet-SSD-TensorRT

TODO:

• FP16 Implementation
• Change Custom layers IPlugin to IPluginExt

The bug has been fixed