ujsyehao/tensorrt-object-detection

tensorrt-object-detection

tutuorial

This project supports the conversion of caffe model to tensorrt model. It includes common IPlugin layer implementation and common errors in conversion.

support detection model

• SSD
• mobilenet-ssd
• mobilenetv2-ssd
• Refinedet

support multiple resolution

• 320x320
• 512x512
• 300x300

support tensorrt version

• tensorrt 4.0.0.4
• tensorrt 4.0.1.6
• tensorrt 5.0.2.6

Iplugin layer description

• detection_output(refinedet used)
• relu6

compile the project

• download the link[https://pan.baidu.com/s/170phX6-q7kj5y9DmFC1yug], copy the caffemodel to model directory
• modify CMakeLists.txt(cuda path, opencv path, tensorrt path)
• run a.sh

TODO

• update code
• add multiple version support