Official YOLOv7 Pose vs MediaPipe | Full comparison of real-time Pose Estimation | Which is Faster?

Поділитися
Вставка
  • Опубліковано 11 січ 2025

КОМЕНТАРІ • 51

  • @LearnOpenCV
    @LearnOpenCV  Рік тому +2

    Get expert guidance, insider tips n tricks and Create stunning images, learn to fine tune diffusion models, advanced Image Editing techniques like In-Painting, Instruct Pix2Pix and many more.
    Join our Kickstarter campaign now! bit.ly/3JYh7A6

  • @LearnOpenCV
    @LearnOpenCV  Рік тому

    📚 LINK TO BLOGPOST: learnopencv.com/yolov7-pose-vs-mediapipe-in-human-pose-estimation/
    ▶ LINK TO YOLO MASTERCLASS PLAYLIST: ua-cam.com/play/PLfYPZalDvZDLALsG9o-cjwNelh-oW9Xc4.html

  • @johncasey434
    @johncasey434 2 роки тому +8

    It was a real pleasure to watch such a clear and concise comparison. Excellent video 👍

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      Glad you liked it @John. More videos incoming!

  • @ChetanAnnam
    @ChetanAnnam 2 роки тому +3

    Awesome comparison, it reduced my work drastically.

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      We felt the same while working with both YOLOv7 and mediapipe that everyone should know about this comparison! Glad you found it useful.

  • @nvmrenh7938
    @nvmrenh7938 2 роки тому +4

    Great video comparisson between Yolov and Mediapipe man, good thing I saw this video in my UA-cam feed.
    +1 Sub 👍

  • @leonidas1983
    @leonidas1983 Рік тому +1

    great explanation! thanks from Argentina

  • @yohanessatria2220
    @yohanessatria2220 2 роки тому +1

    Nice Video! the test on many cases was so helpful!

  • @rahulagiwal4126
    @rahulagiwal4126 Рік тому +1

    Great Video!! Thank you for the super informative video, was looking for the right pose estimation to use for my dance project and this really helped!

  • @leonidas1983
    @leonidas1983 Рік тому +1

    great work, thanks!

  • @nhattuyen1123
    @nhattuyen1123 Рік тому +1

    thank you so much, this video is very helpful

  • @shoghi2547
    @shoghi2547 2 роки тому +1

    I like your sharing. It is clear and easy to understand.

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      Thank you, glad you liked it 😊

  • @siddharthkumar5206
    @siddharthkumar5206 8 місяців тому +2

    Mediapipe does support multiperson detection now

  • @Favourites_Song
    @Favourites_Song Рік тому

    Great Video sir. Thank you for sharing.

  • @nicopetermann1851
    @nicopetermann1851 2 роки тому +7

    Many thanks for this great video! You mentioned that one can use any object detection model for yolo pose - could you elaborate on that? How could one plug in the smallest version of yolov7?

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому +4

      You would need to retrain the network with a different backbone.
      The authors have trained it for the YOLOv7-W6 model. You can train the model using a different yolov7 model. What you would need is a config (.yaml) file corresponding to the smaller model. You can then train the model using the commands given here: github.com/WongKinYiu/yolov7/tree/pose
      I doubt it would give accurate results for smaller models. I would use mediapipe if I don't need multi-person pose estimation.

  • @murcuschimaawaloyi6619
    @murcuschimaawaloyi6619 5 місяців тому

    Very good explanation. Hi Sir. I have been following your tutorial on how to train a custom Yolov5 object detector as I am doing a school project on vehicle detection. I am having an error on training my model. Is it ok if you can help on this please.

  • @maximklechshev6675
    @maximklechshev6675 2 роки тому

    I felt in love with Mediapipe 1 year ago when I worked with facial pose estimation… but YOLOv7 just outperforms it in terms of faces

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      Hi Maxim
      Are you talking about face Detection or Facial Landmarks Detection using YOLOv7?

    • @maximklechshev6675
      @maximklechshev6675 2 роки тому

      Hey!
      I’m talking about Facial Landmarks Detection. I fine-tuned and used ensemble instead

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      Great, do you have a repo you could share?

    • @maximklechshev6675
      @maximklechshev6675 2 роки тому +2

      @@LearnOpenCV I worked with medical sensitive data(

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому +1

      No Issues!

  • @GeoffY2020
    @GeoffY2020 2 роки тому +1

    Hi thanks for the nice job in the video ... I'm doing single image (3 image consecutive) face landmarks alignment, is Yolo better than MP ?

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      Thanks for the kind words Geoff!
      YOLO does not have good enough number of points for Face landmarks alignment. Mediapipe has a dedicated face mesh model that gives 468 3D landmark points on the face. You can check out our blog post on Creating Snapchat filters using mediapipe. You can learn about how to use the different points for your application.
      learnopencv.com/create-snapchat-instagram-filters-using-mediapipe/

  • @rohitghule9437
    @rohitghule9437 Рік тому

    Can we tweek mediapipe to work even when upper part of body is not visible

    • @LearnOpenCV
      @LearnOpenCV  Рік тому

      The pose solution model consists of two models. The detection model (that detects the body), and the landmark model (that maps the landmarks). If you can make the detection model detect the body without its upper part, theoretically, the solution will work.

  • @luchoSwimmer
    @luchoSwimmer Рік тому +1

    nice video

  • @abdurrazzak1612
    @abdurrazzak1612 Рік тому +1

    Excellent

  • @nireksaravanan8532
    @nireksaravanan8532 9 місяців тому

    Are you sure Mediapipe doesn't support Multi-person? pls verify once

    • @LearnOpenCV
      @LearnOpenCV  9 місяців тому

      As of 2024 Jan update, Mediapipe does supports mutiperson pose but limited to 5 at a time.
      For further info check out:
      developers.google.com/mediapipe/solutions/vision/pose_landmarker/

  • @dj.qb91
    @dj.qb91 2 роки тому +1

    What about in images

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому +1

      As mentioned in the summary section, it's better to use YOLOv7 or other pose models as mediapipe is optimized for real-time performance which is more suitable for video inference.
      Hope that helps!

    • @dj.qb91
      @dj.qb91 2 роки тому

      @@LearnOpenCV so with Multiperson which is better than yoloV7.

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому +1

      @@dj.qb91 For Multiperson we're checking out MMPose next -> github.com/open-mmlab/mmpose. You may also check it out and compare with YOLOv7. Check this out for getting started: mmpose.readthedocs.io/en/v0.29.0/get_started.html#inference-with-pre-trained-models

    • @dj.qb91
      @dj.qb91 2 роки тому +2

      @@LearnOpenCV thanks 🙏🏾

  • @nirasinghania6616
    @nirasinghania6616 2 роки тому +2

    👍👍

  • @sanidhyasrivastava4307
    @sanidhyasrivastava4307 4 місяці тому

    Yolo+mediapipe

  • @H_-vy2mz
    @H_-vy2mz 2 роки тому +1

    호호

    • @LearnOpenCV
      @LearnOpenCV  2 роки тому

      I'm not sure what that means, but I'm hoping you liked it! 😊