Home

Welcome to the edgepipes wiki!

Jetson Nano

Converting Models to TensorRT

Clone the repository https://github.com/jkjung-avt/tensorrt_demos.git This repo contains a set of different models that can be converted to TensorRT that is optimised for Nvidia CUDA.

Convert and test a few networks. Then

Convert Face-mask Yolo to TensorRT

Download the weights (yolov3_mask_last.weights ) from the face mask detector made with Yolov3: https://github.com/VictorLin000/YOLOv3_mask_detect

Also download the config file: yolov3_mask.cfg

Rename both files to yolov3_mask_last-416.xxx

Then from the yolo folder in the tensorrt demos: Do: python3 yolo_to_onnx.py -c 3 -m yolov3_mask_last-416

and python3 onnx_to_tensorrt.py -c 3 -m yolov3_mask_last-416

Now you should have a converted file that you can use with the following python3 trt_yolo.py -c 3 -m yolov3_mask_last-416 --rtsp rtsp://192.168.1.237:7447/5c8d2bf990085177ff91c7a2_2