False detection.
BlcaKHat opened this issue · comments
have a lot of false detection, detecting people without helmet as people.
The models also detect people without a helmet, except for detecting people with a various color helmet.
can it be useful for counting the number of people wearing helmet.
It also detects cap with Red,Yellow, White, Blue as helmet.
Also people wearing RYWB t-shirts are also getting detected as wearing helmet
I think -
About counting people,
Line 140 in ssd_pelee.py, Decide your score limit as expected_value
countPeopleWearingHelmet = 0
if score>expected_value and label_name!="N":
countPeopleWearingHelmet=countPeopleWearingHelmet+1
or you can consider changing this parameter than score
conf_thresh
@komalpharate How did you download the pretrained model files. These are uploaded in Baidu cloud which cannot be accessed? @wujixiu Could you suggest something?
have a lot of false detection, detecting people without helmet as people.
Hi,
how do I even run the detection ? Downloaded the model etc, but which script to run ? Can you please help ?