qengineering / face-recognition-with-mask-jetson-nano Goto Github PK

Hi,
I tested it on two cameras :-

QHM495B supports 480P high-definition images and true colour images. Anti-flicker 50Hz, 60Hz or outdoor : 480P
&
Lapcam web camera HD 720P High resolution Video LWC-042  : 720P

On the first one which is 480p, the accuracy is low as the default exposure is low and video is coming dark. Hence its saying 'fake' most of the time.
While on the second one which is 720p, accuracy is really good, video is well lit but the video is slow compared to first camera.

Any reasons as to why this is happening ? And how to solve it.
Id definitely like to use second camera, but how to solve the slow video issue ? Is the video slow because the camera is 720p ?

Hi Sir,

Ive got a .wav file which Im playing whenever any person's Face is detected in camera. Please have a look at the playsound section of this snippet (main.cpp). The sound is playing but I think its slowing down the FaceRecognition also. Is system command slowing down the recognition/detection process ? Is there a faster /better way to play audio without slowing down the recognition/detection ?

#ifdef RECOGNIZE_FACE
        cv::String Str;
        cv::Scalar color;
        int  baseLine = 0;

        switch(obj.Color){
            case 0 : color = cv::Scalar(255, 255, 255); break;  //default white -> face ok
            case 1 : color = cv::Scalar( 80, 255, 255); break;  //yellow ->stranger
            case 2 : color = cv::Scalar(255, 237, 178); break;  //blue -> too tiny
            case 3 : color = cv::Scalar(127, 127, 255); break;  //red -> fake
            case 4 : color = cv::Scalar( 32,  32, 255); break;  //pure red -> mask
            default: color = cv::Scalar(255, 255, 255);
        }

        switch(obj.NameIndex){
            case -1: Str="Stranger"; break;
            case -2: Str="too tiny"; break;
            case -3: Str="Fake !";   break;
            case -4: Str="Mask !";   break;
            default: Str=NameFaces[obj.NameIndex];
        }
        cv::Size label_size = cv::getTextSize(Str, cv::FONT_HERSHEY_SIMPLEX, 0.6, 1, &baseLine);

        /////// ======== PLAY SOUND ============================ ///////////////
        // DETECT NAME & OUTPUT ON CONSOLE LOG
        cout << "NAME DETECTED AS \n\n " << Str << std::endl;

      //PLAY WELCOME SOUND   
          string str1 = "aplay ";
          str1 = str1 + "-Dhw:0,3 Welcome.wav" + " & ";
          const char *command = str1.c_str();
          system(command);
        } 

On the FaceMaskRecognition console im getting this error. I think whenever a face is detected audio is being called to play while the previous audio file hasnt finished playing. Is there a way to solve this ?
aplay : audio open error: device or resource busy

Hi Sir,

Will it be beneficial use use binocular camera, in terms of live face detection and overall face detection?

Keeps throwing "Fake"/"Stranger" even for a person with 3 photographs.
While sometimes it works. This happened when we took the photo in the same room, we tested this application for.
I cant understand why this erratic behavior, is it due to lightening of room or anything else

Can we use this in android by using ndk?

Sir my college has approx 3800 students and some 500 in staff. Will nano be able to handle this much load or shall we test on Jetson Xavier. Or if you can then please suggest an alternative hardware as we are not in college premise right now, but after two months we would like to demo this at our college by being well prepared in advance.

How can I disable the auto database fill feature.
If the face isnt found in the DB, I dont want to add it automatically.
How to do this ?

Hi,

I really appreciate your work and great responsiveness of the issues.

Did you make any test that can we recognize people with mask ? I mean I have both phot in the DB bot maske and normal.

Can system recognize me with mask ? is there any tested accuracy ? Or we need a new model to recognize people wearing mask

Thanks for great repo, I wonder if we can run with GPU or both? Thanks again

Hi Sir,
I want the face recognition to start working only when someone is lets say less than 2 feet near to the camera.
I dont want to recognize distant people. Can you tell me which like to tweak , ill change as per my requirement.

Hi @Qengineering

Program working DEBUG mode but crashing at Release mode...

Retina face is crashing..

really strange..

Hi Sir,

You have 4 models ( RetinaFace, Mask, Arcface & AntiSpoofing ) running over single video.

1.] What exactly is this method of running multiple models known as?

2.] Is there any documentation/book on how to run multiple models at once ?

3.] Why Im asking is because I have a model for License Plate detection and another model for License Plate Recognition, and I want to combine both the models in same way as you have done. (First model tries to locate license plate in video -> If found -> Second Model reads the number plate)

Hi Sir,

Is it possible to install this on normal linux desktop ? Since both paddle and ncnn can be installed on desktop also, hence I suppose it might work on desktop also , right ?

Tried to build in Jetson Nano on Ubuntu 18.04 and this error showed up:

g++: error: /usr/local/lib/paddle-lite/libpaddle_api_light_bundled.a: No such file or directory

I have paddle-lite for Armv8 installed, but there's only a single file inside the lib directory which is libpaddle_light_api_shared.so. As i know, the one with libpaddle_api_light_bundled.a is the x86 version

Hi！
I have successfully installed the environment needed to run the program on Jetson Nano, but I don't know how to compile this project to make it run successfully.

Sir,
Let's say I want to disable face detection and recognition until a certain condition is true( decided by another external code).
How can I achieve this.
At exactly what line do I need to put a boolean condition :
if true : execute Face detection and recognition ,else: do nothing.

Hi buddy,

Awesome work, but I think mask detection isnt working at all. I tried with lots of different masks with 4 students in front of the cam but its always showing "OFF".
I think you might have forgotten to update/enable something in the github code.

Hi Sir,

The code auto-assigns number to each and every person (person10, person 12 etc).
Then whats the logic behind calling someone a stranger ?

I think sometimes the image descriptors are ZERO and the code is trying to access this UNINITIALIZED BUFFER,hence its crashing sometimes at 30 seconds and sometimes after 10 minutes.

Theres one small issue. When the algorithm detects a tiny face (personA) at a distance, and then suddenly if someone comes near the camera (personB), then personB is also recognized as "too tiny".
But when the personA from behind goes away, then personB is aptly recognized.