Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Pospielov Serhii
    @pospielov
    Not sure, I think the most resource-consuming are CPU and GPU here
    Mohamed Essam
    @3oos97
    Hi Everyone, i have used python sdk to communicate with Ip camera with default compreface that contains about 300 subjects with 4 photos for each one, i used flask app to render the live stream of the camera on a web page but unfortunately the stream is lagging very much, how can i improve the performance of the live streaming?
    Allen5017
    @Allen5017
    Hello. I defeated it only by processing frames from a separate thread ;)
    1 reply
    Allen5017
    @Allen5017
    я писал скрипт на python
    I was writing a python script
    steppel
    @steppel
    Hey, all. I need help :-( .. running Unraid and compre-GPU, A4000 with nVidia driver installed. Created a test app and tried a test image. Nothing happens except an error in the log:
    2022/04/12 08:41:09 [error] 304#304: *1 upstream timed out (110: Connection timed out) while reading response header from upstream, client: 10.253.1.2, server: ui, request: "POST /api/v1/recognition/recognize?face_plugins=landmarks&face_plugins=gender&face_plugins=age HTTP/1.1", upstream: "http://127.0.0.1:8080/api/v1/recognition/recognize?face_plugins=landmarks&face_plugins=gender&face_plugins=age", host: "10.10.0.110:8800", referrer: "http://10.10.0.110:8800/test-model?app=a6f68ac4-f66d-4285-846f-c221c3c1cd52&model=b7935299-3929-4c5a-9317-c83e7bc9943f&type=RECOGNITION"
    Docker is running in bridge mode
    steppel
    @steppel
    Oh and rhe regular compreface is running without issues, i just cant get the GPU version to run
    Pospielov Serhii
    @pospielov
    Unfortunately, your GPU fails with the same issue as all Nvidia 3xxx GPUs. Here is the discussion:
    exadel-inc/CompreFace#755
    we need to update CUDA, so it will start working.
    steppel
    @steppel
    Interesting, tganks! Will keep an eye out for updates.
    Rinaldi Espera
    @rinaldi.espera:matrix.org
    [m]

    Installed. All 5 containers running based on docker ps. But from UI, when demo app is clicked, I am getting "Error during synchronization between servers: compreface-core executing GET http://compreface-core:3000/status". There was similar reported case and suggested to wait (but waited and been awhile already), and also suggested that compreface-core isn't running (but docker ps says it's already up). Last lead suggested that CPU does not support AVX (checked the specs and confirmed supported). Looking at the compreface-core logs, the only line which seems being highlighted is "!!! no internal routing support, rebuild with pcre support !!!". Other than this, nothing obvious which suggests any lead on the issue.

    Another thing I noticed, which may be unrelated, but just observation, from docker ps, compreface-core goes back "Restarting" status unlike the rest of the containers. And seems to stay in that status for awhile, then goes to "Up" then goes back to "Restarting" again as suggested by the text "...n seconds ago".

    Appreciate any lead/help in the issue. Thanks.

    Allen5017
    @Allen5017
    It seems I had it like this when my laptop didn't support avx
    1 reply
    Pospielov Serhii
    @pospielov
    Could you upload the logs somewhere and send the link here?
    1 reply
    Pospielov Serhii
    @pospielov
    Hi @/all, we just released a new version of CompreFace - 1.0.0
    You can find the changes and install instructions on GitHub:
    https://github.com/exadel-inc/CompreFace/releases/tag/v1.0.0
    Anish703
    @Anish703
    Hello, is there any eta for the support of RTX 3XXXX?
    1 reply
    nagem07
    @nagem07
    Hi, I was going through your documentation and I saw it is advised to run two servers for production environment. Could you advise at to how to setup each? I see data integrity is maintained through PostgreSQL notifications, is there anything I need to setup for this to happen?
    3 replies
    vijayjshah
    @vijayjshah
    Hi , Could you suggest some VMS ( Video Management Software ) for integration of CompreFace ?
    1 reply
    Anirudh Nimmagadda
    @ack_inc_gitlab
    Hey guys, does CompreFace's admin server have a documented API?
    1 reply
    Anish703
    @Anish703
    Would it be possible to write each individual applications data to different tables in the db?
    2 replies
    Allen5017
    @Allen5017
    Hello, please, can someone suggest? How do you receive and send frames to CF if there are several cameras? Are you working in multiple processes, threads? How to do it in a stable way? I have +- one recognition with one camera. Now I have to do the second one, but it's difficult… I want to keep the frames black, to improve the quality of recognition and motion analysis. Now I process only the current frames, as CF manages, and the rest of the frames pass by… Can someone guide me?
    "Queue"
      • black
    nagem07
    @nagem07
    Hello again,
    I would like to train my own custom model based on insightface, I am assuming the implemented model is based on Pytorch?
    Also, I see on your documentation it is advised to upload the model to Google drive before modifying the Calculator class. Could you advise how the new model should be added as I see four values for each models, which is the model name, some encrypted string, a tuple of two integers and lastly an integer which is either 200 or 400.
    1 reply
    Allen5017
    @Allen5017
    Hello. Maybe someone has already made an assessment of the position of the head according to landmarks landmarks2d106? I need to determine if there is a face in front of the camera or not. Face direction
    martinenkoEduard
    @martinenkoEduard
    Is there a way to use CompreFace in offline mode? What account should I use then?
    3 replies
    Is it possible to use GUI in this mode?
    martinenkoEduard
    @martinenkoEduard

    can anyone explain to me how to change UI a little bit.
    I mean what software do I need to compile it? (webpack and so on)

    I really like the UI, but want to change it a little bit to use in my project!
    And I am noob in frontend.

    Pospielov Serhii
    @pospielov
    We have instructions for contributors. They are actual for frontend: https://github.com/exadel-inc/CompreFace/tree/master/dev
    The idea is that you run all CompreFace servers, but the frontend is compiled on a host.
    Alternative - just build every time CompreFace using docker-compose build from dev folder. It's the easier way, but much longer to develop.
    nagem07
    @nagem07
    Hi, I've been trying to reach the Exadel team through support options on their website but up to now no one has gotten back to me. Could you please provide the best means of getting in touch?
    5 replies
    gale952
    @gale952
    How to implement the CompareFace in my bionic human head face recognition software?
    1 reply
    gale952
    @gale952
    Thanks for the reply, Serhii.
    gale952
    @gale952
    I am building a human head wit capabilities to recognise a human face and perform chat with the person recognised. The head is stand alone and all software maintenance is done remotely from PC via WiFi local network. It is possible to connect a keyboard and monitor to the head and perform some maintenance locally as well. The head vision software should recognise already seen person which image is stored in local disk or save new person face snapshot into the database for the later recognition. The head has 2 DOFs (up-down/left-right) in order to follow the persons movements. When a person is recognised or new one "met", the conversation shall be triggered. Separate software module is responsible for the chat (basically it is an intelligent not Internet dependant chatbot). All head software should work autonomously and standalone. Hope this description is enough for the beginning. I would be very happy if CompreFace software is right framework for the head's vision task.
    1 reply
    gale952
    @gale952
    I was going to use Khadas VIM4 Amlogic A311D2 SBC which does not support AVX. Perhaps I should use one of the Intel's NUCs but it is far more expensive. By the way, I do not know why I do not see your message in the common chat.
    Pospielov Serhii
    @pospielov
    I usually reply in threads, you can find it under your messages.
    You can use other boards with AVX support. As far as I know, there are even AMD-based boards with AVX support. But yes, unfortunately, they are much more expensive than Khadas VIM4
    Rodrigo Garcia Scara
    @rodrigo_ags_twitter

    Hi there!
    Congrats for your great work!

    Do you have an idea of ​​the minimum requirements that an image must have to be acceptable?

    Resolution, distance in px between eyes, illumination, etc?

    1 reply
    Im getting a lot of false positives with confidence around 95/99, in a set of 200 images. I think that the quality of pictures is not enough.
    Thank you!!
    gale952
    @gale952
    I'am not at that stage of project yet!
    martinenkoEduard
    @martinenkoEduard

    Hi there!
    Congrats for your great work!

    Do you have an idea of ​​the minimum requirements that an image must have to be acceptable?

    Resolution, distance in px between eyes, illumination, etc? |

    I have the same issue. With low quality images there are a lot of false positives.
    Maybe there is a guide on how retrain the model to fin particular data?

    1 reply
    Btw how do I start 2 or maybe more more detection models? I want to improve performance, by using RAM.
    1 reply
    I mean how do I increase simultaneous requests?
    Slaus Blinnikov
    @SlausB
    Hello! In our service we want to check if user provided his correct unedited image. Does ComperFace provides a function to detect image integrity? So to check if image wasn't photoshoped or something.
    Thanks in advance.
    1 reply
    Mikhail Shashkov
    @shashkovma_twitter
    Hello everybody!
    I would like to share with you the announcement of the hackathon in which CompreFace is participating. Feel free to join our Discord server, if you have any questions, thanks.
    Mikhail Shashkov
    @shashkovma_twitter

    Is Open Source an integral part of your life? Or are you just getting to know it?
    Either way, join the OpenHack contest by Devexperts to contribute to the development of Open Source, and сompete for a € 10,000 total prize fund!

    OpenHack is a hackathon with a multitude of open-source projects to choose from. Developers, designers, QA engineers, technical writers – we have challenges for everyone!

    OpenHack will last throughout October and will take place online. It’s a great opportunity to improve your skills and communicate with other open source enthusiasts. Become part of the global community!

    How to participate:

    • Choose an interesting project and tasks,
    • Work independently or create a team,
    • Find the solution and implement it.

    Feel free to join our Discord Server (https://discord.gg/xw4U4nCCkh) and ask anything, we’ll be there to help you!

    Registration has already started, join via link by September, 28: https://bit.ly/3qGd6Ff

    Komron Juraboyev
    @arab04:matrix.org
    [m]
    1. When we compare two different faces it returns high similarity in face verification REST API what is the reason for that?
    2. How many points does it uses when it makes face verification?
    3. I saw that there are 5 main points and does that mean that it puts only 5 points to the face when it verifies user?
    4. what type of algorithms does it use when it verifies user face?

    I really need answers to all this questions, I would be very glad if someone could give me explanation to my questions

    Thanks in advance

    1 reply
    Trần Minh Chiến
    @ChienLady
    I have a question: How can i find the status (coef0, coef1) of my arcface model ?
    https://github.com/exadel-inc/CompreFace/issues/910#issue-1391716768
    Saran Laoyookhong
    @saran.rez_gitlab
    Is there a way to reset my login password?
    Mikhail Shashkov
    @shashkovma_twitter

    Hey everyone!

    I'd like to tell you that today is a deadline for joining OpenHack hackathon. You can participate online at a convenient time. (CompreFace is already with us!)

    Contribute to the development of Open Source and сompete for a € 10,000 total prize fund!

    How to participate:

    1️⃣ Choose an interesting project and tasks,
    2️⃣ Work independently or create a team,
    3️⃣ Make at least 3 merged PR's and get a guaranteed prize.

    Feel free to join our Discord Server (https://discord.gg/xw4U4nCCkh) and ask anything =)

    🔗 Join OpenHack via link: https://bit.ly/3TclZDt

    Wisdom
    @wisdompython

    Hello everyone, Compreface is very cool,

    I need some help though, i am currently using compreface for just its verification functionality

    The main issue is it is consuming a lot of RAM

    is there a way i can reduce its consumption

    Pospielov Serhii
    @pospielov
    Hello, answered here: exadel-inc/CompreFace#946
    nagem07
    @nagem07
    Hi, I understand that Euclidean distance is converted through this formula
    (tanh((first_coef - distance) * second_coef) + 1) / 2
    to return similarity score. How are the coefficients calculated?
    1 reply