Acknowledgement: This repo is an inherited from this amazing repo. Thanks
Make sure right the path of config, frozen and label file
I write this code suitable for all ROS1 and normal run. dst_new is use for normal run, It will read your camera on laptop and if image contains human, every 2 seconds it will save image and its label follow VOC formal in .xml file all in data
Second, for ROS, dts will subscribe to a topic contains image which had configured in sub_node, everytime this topic receive image and have human, it will do the same. I do this for semi-sampling image human use for my human detection custom project.
Here is a sample that I take myself 🇻🇳 😄 🇻🇳. Detector just detect human, the confident around 0.6-0.8
File convert can convert stream image in a topic into video.