Issue Regarding Safety jacket Detection, number of people and violations

[mohitmathur95]

Hi,
I ran the algorithm on a video (screenshot attached in absence of rights to upload video here). The algorithm is not able to correctly count the violations of safety jacket, is not correct with detecting number of people. It works correctly for some videos and doesn't work correctly for a few. What could be the possible reasons. Any suggestions to take care of this?

[asm128]

You forgot to mention that it doesn't detect anything about the helmets. I see people without helmet with a square on its head and then people without jacket with a square on the chest. Like there is no difference between one and the other:

Sadly, this is not the only thing from Intel that doesn't work. When I installed OpenVINO it points to some tutorials about face detection and car detection. None of those are even able to build and I had to open issues for those too. It seems their development processes prevent them from keeping things updated and working and that all the hassle comes from incompatibilities between the things they provide (even if they link one to the other as if the samples were for that version).

So, I'm moving on to some other solution. I'll be back around 2030 because I predict all this will take a long time to be fixed.

[KnightRan]

Dear gamesgomez123 ,
Good day to you, may I know what Openvino version are you running to run this application .I have the latest Openvino_2021.3.1 but it didn't worked well with that version. Thank you

[KnightRan]

You forgot to mention that it doesn't detect anything about the helmets. I see people without helmet with a square on its head and then people without jacket with a square on the chest. Like there is no difference between one and the other:

Sadly, this is not the only thing from Intel that doesn't work. When I installed OpenVINO it points to some tutorials about face detection and car detection. None of those are even able to build and I had to open issues for those too. It seems their development processes prevent them from keeping things updated and working and that all the hassle comes from incompatibilities between the things they provide (even if they link one to the other as if the samples were for that version).

So, I'm moving on to some other solution. I'll be back around 2030 because I predict all this will take a long time to be fixed.

Same goes with me, as the system still detect the helmet even i am not wearing it on camera stream.