I was running the sample "safety_gear_detection" program in the dev cloud and I observed the wrong inference. Is because training data was not correct. How I can train the model in the program to provide accurate results. In one scenario "Hard Hat" is not detected and in the second scenario, it detected "Hard Hat" although it's not there.
It looks like the confidence values are low in the sample output so to achieve a better, more correct result you would likely need to re train the model. Please let me know if you have any further questions.