30+ FPS 3D Object Detection with MediaPipe and OpenCV: The Latest Breakthrough in CPU Processing

Поділитися
Вставка
  • Опубліковано 25 гру 2024

КОМЕНТАРІ • 45

  • @NicolaiAI
    @NicolaiAI  Рік тому +1

    Join My AI Career Program
    www.nicolai-nielsen.com/aicareer
    Enroll in My School and Technical Courses
    www.nicos-school.com

  • @arpitkapil8932
    @arpitkapil8932 3 роки тому +1

    Thank You for this amazing video.......U don't know what u do for me.....this is the reference which i can use for my college project of Vacant Chair Detection

    • @NicolaiAI
      @NicolaiAI  3 роки тому +1

      Thank you very much, really appreciate it and glad that i could help!

  • @sarpermelikertekin5386
    @sarpermelikertekin5386 3 місяці тому

    Do you have a repo for that, it looks great

  • @candikahandaru8251
    @candikahandaru8251 10 місяців тому +1

    can u do it as custom dataset. maybe like tree detection that can count and calculate diametre of the tree

  • @outragepanda830
    @outragepanda830 Рік тому

    Hey great video. I just have one suggestion. At the star of the video show the working model which may help us to understand more while you are coding

  • @anshumannath7400
    @anshumannath7400 3 місяці тому

    Has the object detection been deprecated by Mediapipe. It is taking way too long for executing this in jupyter notebook

  • @axboltx
    @axboltx Рік тому +1

    Is there a way to detect vehicles while using opencv ?

  • @cristianlazoquispe4472
    @cristianlazoquispe4472 2 роки тому +1

    great work 13:29

  • @Burubiri
    @Burubiri 8 місяців тому

    can use for object measurement?

  • @yifeixue1628
    @yifeixue1628 3 роки тому +1

    Thank you very much for your contribution, however, I just found this method only can be used for shoe chair cup or camera, wish one day can use this in a wider field

    • @NicolaiAI
      @NicolaiAI  3 роки тому

      Thank you very much! Really appreciate it and I'm sure they will update it over time to more objects or make it possible to train a model urself

  • @totem0988
    @totem0988 Рік тому

    Hi, is there a way to run this using GPU in browser?

  • @markopopoland
    @markopopoland 3 роки тому

    Sir I'm going to detect wooden parts on gripper or convoyer? Can I use medialipe for this purposes ? Need to find pose of part ? How can I train parts for mediapipe ?

  • @leilaghasemzade5880
    @leilaghasemzade5880 Рік тому +1

    Thanks for the tut. I need to use mediapipe object detection in python. I cannot use yolo or other things, only mediapipe. I cannot use other languages, only python and I don't need 3d obj-det. Is there anyway to use mediaipe object detection solution in python? if no way, I can use 3d obj-det instead of 2d obj-det but is there any drawback/disadvantages?

    • @NicolaiAI
      @NicolaiAI  Рік тому

      I dont Think they have any general 2d object detection model. What is the reason u cant use the best object detectors out there?

    • @leilaghasemzade5880
      @leilaghasemzade5880 Рік тому

      @@NicolaiAI Firstly, I have several algorithms that they use mediapipe as the basis. I want my object detection module use medipipe as well to make all of the uniform. second, the most important reason is speed/FPS. previously I was using Yolov3 but the FPS was very low, not satisfying so I decided to use mediapipe obj-det which is very light-weighted (CPU based?) and fast.

    • @leilaghasemzade5880
      @leilaghasemzade5880 Рік тому

      @@NicolaiAI Hi Nicolai, waiting for your response :)

  • @bardiamojra9997
    @bardiamojra9997 3 роки тому +2

    Im second! Thanks for the great tutorial! Any chance you could cover point cloud to mesh grid and key point feature extraction. All the tutorials are in matlab, it makes me sad lol

    • @bardiamojra9997
      @bardiamojra9997 3 роки тому

      honestly never mind, this is great. I think I can figure it out, I will look into mediapipe module

    • @NicolaiAI
      @NicolaiAI  3 роки тому

      @@bardiamojra9997 thank you very much. I'll try to look into it when I have time!

  • @0GRANATE0
    @0GRANATE0 2 роки тому +1

    really cool! How can you train your own model to recognize your custom object?

    • @NicolaiAI
      @NicolaiAI  2 роки тому +1

      Thanks a lot for watching! I have a video about custom 2d object detection from scratch with a yolov5 model here on the channel and how to deploy it.

    • @0GRANATE0
      @0GRANATE0 2 роки тому +1

      @@NicolaiAI Thank you! I'll watch that! But how do you do it with a 3d object? I know you can create a 3d Model of that custom object and then render it from all sides and colors and shades... but how do you know the camera perspective? do you also add those information to the training? hmm

    • @NicolaiAI
      @NicolaiAI  2 роки тому +1

      I guess u would just train it to do rough estimations. It's rather complex if u have to make it camera dependent I think

    • @rindtier7287
      @rindtier7287 Рік тому

      @@NicolaiAI anything new about this?

  • @matthewharrison3813
    @matthewharrison3813 8 місяців тому

    Thanks for the brilliant clear video. If you ever revisit this topic, I'd be interested in some coverage of the accuracy of the results. Is this method suitable for translating to the real world? With appropriate translation between the camera frame and the world frame, would the 3d co-ordinates given by this model match the detected object's position?

  • @nguyenworld21
    @nguyenworld21 2 роки тому

    Do this can help detect vehicle or car ? I want to use it for parking lot detection so I have to detect vehicles first

    • @collinb5524
      @collinb5524 Рік тому

      No this with this dataset u can only detect like 7 objects, its very limited

  • @Yukki7934
    @Yukki7934 3 роки тому +1

    Hey, Thanks for the wonderful explanation, I just ran over a issue. INFO: Created TensorFlow Lite XNNPACK delegate for CPU. and the Camera works but the Model doesn't shout. Thanks in advance .

    • @NicolaiAI
      @NicolaiAI  3 роки тому

      Thank you very much! Is it an error or a warning. I recognize that message but thinks it comes Everytime u use tensorflow

    • @roopeshk.r3219
      @roopeshk.r3219 3 роки тому +1

      @@NicolaiAI Hi Code Lib, Thanks for the excellent video ..i also got this Warning ,,,,After that the Model was not showing, how to resolve this?

  • @soltified
    @soltified 3 роки тому

    goodwork man

  • @omkaryadav-ji3um
    @omkaryadav-ji3um 3 роки тому

    Can you help for car parking detection through camera

  • @wilbsmond
    @wilbsmond 2 роки тому

    Ah this is super cool! This kind of 3D bounding box with 3D orientation vectors is what I'm envisioning for my custom detector. I have my trained custom 2D object detector. Do you know how (if possible) to use mediapipe to do something like this with a custom 2D detector?

  • @유영재-c9c
    @유영재-c9c Рік тому

    hello. I would like to get advice from someone who has experience with object detection in the form of 3D cuboids. I had an opportunity to experiment with 3D bbox object detection in autonomous driving, but I had no knowledge at all. It also covers basic input/output formats and base models. It seems like there is less content compared to 2D detection. Any site where you can study the content or a framework that you can recommend would be great. If you know anything about the related content, I would like to share it with you. thank you!

  • @융코맨
    @융코맨 2 роки тому

    Can I detect other model with using mediapipe?(not a cup, chair, camera)

    • @ahmedtalaat4886
      @ahmedtalaat4886 11 місяців тому

      Hey man hope you found a way to do that, If yes please can you tell me how

  • @shekharkumar1902
    @shekharkumar1902 2 роки тому

    Thanks for nice video , unfortunately it just worked on 'Shoes' and 'Cup' model and its also not working very smooth . Though Objectron dataset have other classes like bike ,book, bottle, camera, cereal box, chair, cup laptop, shoe. I tried it on bottle and book , it didnt work. :-(

  • @wishIKnewHowToLove
    @wishIKnewHowToLove 3 роки тому +2

    First