This function I wrote performs the image and txt saving process. You can also use this function on a model you have trained before. If your model works well enough, it will easily detect the images and save the labels. You can edit the saved data and train with your new data quickly.
first of all, tensorrt transform using this repository of the yolo weights you have trained
https://github.com/jkjung-avt/tensorrt_demos
You should edit yolo/obj.names according to your project class names
And You should change plugins and utils your tensorrt_demos file
The input parameters can be changed using the command line :
python3 img_test.py -i <input_images_file> -w <trt_weight_name> -t <image_type(jpg,png...)>
For running with custom weights :
python3 img_test.py -i images -w yolov4-tiny -t jpg
To display the help :
python3 img_test.py -h