Template Matching.

[sweetngx]

Actually, it's exactly the project I'm looking for, but it doesn't exactly serve my purpose. Can a similar change be made to the template match feature as in opencv? Is it possible to do this?

Is it possible to make changes to this project? I have enough budget, can you customize this script?

[norihiro]

Can a similar change be made to the template match feature as in opencv? Is it possible to do this?

The code itself is partially ready to implement other detection algorithm.
Currently dlib-based face detection is hard-coded here but it can be replaced with other implementation.

obs-face-tracker/src/face-tracker-manager.cpp

Line 24 in 486ade5

detect = new face_detector_dlib();

can you customize this script?

I'm working on this repository at my spare time. It depends how much quality you want to have and how many days I can take.

[sweetngx]

@norihiro Thank you for the quick turnaround. I'm not really good at coding, but I don't think anyone else can do the project I dreamed of...
I have no idea how to even create the .dat file for the object I want. At the moment, I only use a script with opencv template match feature, which is written with python, but it is very cumbersome and I use many intermediary applications to transfer it to obs. Actually, what I want to do is to censor or cover or blur content that I don't want on a camera or media source. So I think some radical changes should be made in the project. Maybe it is wrong to ask you for help here, but there is no other address that I can reach. Maybe you don't do tailor made applications, but I really need it and the budget is not a problem at all. If you are interested, you can contact me at bilgedondurma35@gmail.com for all the details about my project.

[Phoenix616]

I'm actually looking for something in a similar usecase (automatically bluring faces of people (I assume the head detection will be more important there)) and seeing as this is able to draw markings for detected faces it shouldn't be too hard to add but I don't think it matches the goal of this plugin well. Do you think something along that line could be added here or should I look into a different project/full fork of this detected to censoring?

[norihiro]

For live streaming, I'm not sure that users can allow a situation that a face is revealed when the detector cannot detect the face. Is there a use-case that accepts failure of blurring the face?
Especially, when your face is partially caught by the camera or shooting from side, failure would happen frequently.

[Phoenix616]

I realise that it would of course not be ideal and a warning should be added to the settings regarding that but imo it would still have some benefit as at least a good chance is there to not have a face show at all.

In my case this is more meant as a first pass to at least put some thought into protecting people's privacy directly when streaming (where the situation regarding image rights is often not as clear cut as when publishing records) which then could also allow a simpler manual review of the recording for publishing. (E.g. by cutting out parts or manual bluring faces from the stream recording or ideally direct publishing if no issue is found)

[maxesisn]

For live streaming, I'm not sure that users can allow a situation that a face is revealed when the detector cannot detect the face. Is there a use-case that accepts failure of blurring the face? Especially, when your face is partially caught by the camera or shooting from side, failure would happen frequently.

Through the multi-face detection capability of this plugin, the real-time facial blurring feature will be useful in scenarios like outdoor live streaming, such as showing outdoor scenery while protecting the privacy of passers-by, whether or not it can achieve a 100% detection rate is not the most important (of course I also think that relevant warnings should be added)