Object Tracking and counting using YOLOv8 and DeepSORT | Counting Persons

Поделиться
HTML-код
  • Опубликовано: 20 ноя 2024

Комментарии • 106

  • @ashimasingla103
    @ashimasingla103 9 месяцев назад

    Dear Aarohi
    Your channel is very knowledgeable & helpful for all Artificial Intelligence/ Data Scientist Professionals. Stay blessed & keep sharing such a good content.

  • @ashimasingla103
    @ashimasingla103 Год назад

    Hello Aarohi
    Your channel is very knowledgeable & helpful for all Artificial Intelligence/ Data Scientist Professionals. Stay blessed & keep sharing such a good content. Your channel really needs more likes & share so to reach maximum AI professionals who can encash from it

  • @pifordtechnologiespvtltd5698
    @pifordtechnologiespvtltd5698 9 месяцев назад +1

    Amazing work done. Keep it up

  • @ArbaazShaikh-y3t
    @ArbaazShaikh-y3t 9 месяцев назад

    What I observe is You are first detecting and then incrementing the counter actually, the counting should be solely dependent on each frame not on incrementing a new object every frame caz the object detection model can generate some false detections as well. and the counter may increment. and the actual people in a video are less but the counter can be greater .

  • @Wetu_Vexo
    @Wetu_Vexo Год назад +1

    Thank you for your video ma'am. it's straight to the point. Please have a great day!!

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      You're very welcome! I'm glad you found the video helpful and to the point. Have a fantastic day too!

    • @cattnation6257
      @cattnation6257 6 месяцев назад

      @@CodeWithAarohi need a help i need to track a poor video quality video lost of variation in it i try deep sort but how can i enhance it more any though about clustring and anything ??

  • @Sunil-ez1hx
    @Sunil-ez1hx Год назад

    Hello Ma’am
    Your AI and Data Science content is consistently impressive! Thanks for making complex concepts so accessible. Keep up the great work! 🚀 #ArtificialIntelligence #DataScience #ImpressiveContent 👏👍

  • @soravsingla8782
    @soravsingla8782 Год назад

    Hi Aarohi, your content is excellent and your channel is one of the best Artificial Intelligence channel but still not getting that much of likes which your channel deserves. Hope you succeed #AI
    #ArtificialIntelligence
    #DataScience
    #EducationalContent

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      Thank you so much for your kind words and support! It means a lot to me. 😊🙏

  • @arnavthakur5409
    @arnavthakur5409 8 месяцев назад

    Worth watching again ma'am

  • @arnavthakur5409
    @arnavthakur5409 Год назад

    Explained in a very nice manner, thank you man

  • @Tsukuyomi95
    @Tsukuyomi95 3 месяца назад

    That was amazing work, madame! ❤️ I just have a few questions. What modifications are needed in this code to count multiple objects, not just people? Also, how can I make it count in real-time? Thanks for answering!

    • @CodeWithAarohi
      @CodeWithAarohi  3 месяца назад +1

      If you will notice in github repo. I have this line where I have mentioned the classes = 0 which means only person class. Remove "classes = 0" from this line and it will detect and track all objects.
      results = model(frame, device=0, classes=0, conf=0.8)

    • @Tsukuyomi95
      @Tsukuyomi95 3 месяца назад

      @@CodeWithAarohi thanks for much for answering ❤️ and what i need to modify to count in real time not only in videos?

  • @sandovalon88
    @sandovalon88 4 месяца назад

    Nice Job, is there a way to implement with Home Assistant?

  • @soravsingla6574
    @soravsingla6574 Год назад

    Very well explained

  • @Sunil-ez1hx
    @Sunil-ez1hx Год назад

    Very nice video

  • @prasannapm3220
    @prasannapm3220 9 месяцев назад +2

    Maam how to do class wise counting ?

  • @rafaelcaballosbejano8501
    @rafaelcaballosbejano8501 Год назад

    Hi.
    Thank you for your great channel and so interesting resources. Maybe it sounds like a silly question, but where is the folder with the pictures you use in this video.
    Regards
    Rafael

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      That I didn't upload but you can create a folder and put some images in it and test.

  • @陽明交大-高明秀
    @陽明交大-高明秀 Год назад

    Hi Ma'am, Can you explain more about the deep_sort_weight?

  • @vigneshvicky6720
    @vigneshvicky6720 Год назад +4

    U didnt teach any theory about yolov8 's architecture loss function how it works ext... Directly going to code is not a big think we too get plenty of codes from net we finding difficult to get the theory behind yolov8 plz focus on that. This is my humble request.

  • @philtoa334
    @philtoa334 Год назад

    Nice.

  • @surflaweb
    @surflaweb Год назад

    Thanks dear I will test it

  • @nurmuhimawann
    @nurmuhimawann 9 месяцев назад +1

    Hello ma'am, can object tracking and counting methods like deepsort only be used for video input? what about image input? can it still be used? and how effective is it?

  • @juanmora3756
    @juanmora3756 Год назад

    Hi thanks for the video. One question, can I use yolo-nas instead of yolov8?

  • @muhammadislam9097
    @muhammadislam9097 Месяц назад

    Could you please share the pre-trained model!

  • @breno9138
    @breno9138 5 месяцев назад

    Hello ty for the content

  • @maniac610
    @maniac610 4 месяца назад

    Can you please share the path from where i can download this video and photo.

  • @atulakella4402
    @atulakella4402 Год назад

    Ma'am, how can we change the input to real time and show the real time results in another window real time?

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      Use source as 0 if you want to use webcam, And if you want to use other source then provide that in source.

  • @V.Ajantha
    @V.Ajantha 6 месяцев назад

    Hai mam. I have a yolov8 detection model for my dataset. in that i have five classes. i need to track and count the defective product and i need to reject that defective product from conveyor belt. so for that i need to create a text file with all defective product is assigned as 0 and defect-free product as 1. how to do that.

  • @GauravSingh-zt4us
    @GauravSingh-zt4us 6 месяцев назад

    how to convert these into cuda tensors please update mam

  • @i22-007MUsmanMalik
    @i22-007MUsmanMalik Год назад

    why haven't you used bytetrack?
    and which is better yolov8 tracker, deepsort or bytesort as there is alot of object detection in the video and I am using yolo v8 detector trained on custom dataset.

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      No reason. This topic is requested by somebody, so did a video on it but try to work with bytetrack

  • @siddardhaannapureddy8603
    @siddardhaannapureddy8603 10 месяцев назад +1

    Madam, From where did you get that deepsort files from ? can you please provide me the resources. i know i can clone the deepsort folder directly from your git account but i need the source from which you had downloaded. Thank you in advance Madam.

    • @CodeWithAarohi
      @CodeWithAarohi  10 месяцев назад

      I picked that code long time ago from github. Not sure which repo. Sorry

  • @cattnation6257
    @cattnation6257 6 месяцев назад

    reidentification??????????? without it that code did not do good for my project

  • @oykuecekoken3350
    @oykuecekoken3350 Год назад

    Hello, thank you very much for the video, it was really helpful. However, I am running my code on Jupyter Notebook and I am receiving a 'dead kernel in Jupyter' error. What should I do?

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      You can restart the kernel- try "Kernel" > "Restart & Clear Output"
      If the above solution doesn't work then Close the Jupyter Notebook tab in your web browser, stop the Jupyter Notebook server (you can do this in the terminal/command prompt where you started it), and then restart the Jupyter Notebook.
      Excessive memory usage in your code can also lead to a dead kernel.

  • @ityn07
    @ityn07 8 месяцев назад +1

    Hi ma’am! How did you get the class name of the track id?

    • @CodeWithAarohi
      @CodeWithAarohi  7 месяцев назад

      I am getting it from detection model

    • @ityn07
      @ityn07 7 месяцев назад

      @@CodeWithAarohihow so? At 7:04, it can be seen that the arguments passed to tracker are only bboxes_xwyh, conf, og frame?

  • @alscreation858
    @alscreation858 11 месяцев назад

    video is created but there is no data :- code executed successfully and output video also created but it has 0kb data and can't open.

  • @Tsukuyomi95
    @Tsukuyomi95 3 месяца назад

    Hi Ma'am,
    I'm back with another issue. My code executed successfully and an output video was created, but the video file is 0 KB and cannot be opened. Could you please help me resolve this? ( Btw I use Google colab)
    Thank you!

  • @deshmukhtejas-jg9tb
    @deshmukhtejas-jg9tb 6 месяцев назад

    hi mam . after processing few frAMES ITS GETTING DEAD . I AM USING AWS PYTORCH KERNEL .

    • @CodeWithAarohi
      @CodeWithAarohi  6 месяцев назад

      Check if you are running out of resources

  • @livechemistryfunda1865
    @livechemistryfunda1865 7 месяцев назад

    if we want to only track the person at left and the person at right, what should we do?

    • @CodeWithAarohi
      @CodeWithAarohi  7 месяцев назад

      Calculate the center point of the image and compare the horizontal positions of each detected person's bounding box relative to this center point. This will help you determine which person is on the left and which one is on the right.
      Then filter out all persons except the one on the left and the one on the right.

  • @rajshekhar1036
    @rajshekhar1036 9 месяцев назад

    Can it be done for real time , like using our own webcam for counting ?

  • @salmabif
    @salmabif Год назад

    How can we apply to detect n count objects in images ? kindly guide me

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      Just provide the source image instead of video.

  • @jagdishlamba72
    @jagdishlamba72 2 месяца назад

    How to handle live stream because deepsort create delay in processing and could create overall delay in output

    • @CodeWithAarohi
      @CodeWithAarohi  Месяц назад

      Instead of processing every frame in the stream, process every 2nd or 3rd frame to reduce the computation load. This helps to catch up with real-time streaming.
      Also, Use lower resolution frames to reduce the computation load. You can downsample frames and still maintain reasonable detection performance.
      YOu can try to work with ByteTrack which is lighter than DeepSORT.

    • @jagdishlamba72
      @jagdishlamba72 Месяц назад

      Have you any sample code for byte track

    • @jagdishlamba72
      @jagdishlamba72 Месяц назад

      Thank you very much

    • @CodeWithAarohi
      @CodeWithAarohi  Месяц назад

      @@jagdishlamba72 ruclips.net/video/Jkf9t6jG3vg/видео.html

  • @shivangichaudhary2262
    @shivangichaudhary2262 Год назад

    Hi, ma'am , I have my costumed trained model of Yolov8 using CLI how can we use DEEP SORT for tracking purpose in CLI ? I mean how can we use deep sort algorithm using CLI ?

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад +1

      You can create a .py file and write a code to call deepsort algorithm there. And then run that file from cmd.

    • @shivangichaudhary2262
      @shivangichaudhary2262 Год назад

      Thank You ma'am@@CodeWithAarohi

  • @pushpendrakushwaha604
    @pushpendrakushwaha604 Год назад

    I am working on a project that is sapling counting using a drone. I am using Yolo V8 for object detection. The issue I am facing is that when the drone takes a U-Turn to the next area of the field basically other lines of saplings the previous saplings are counted again in that camera rotation. Can you suggest any idea how to figure this out? I want to do something that will help to stop that extra counting of that sapling while the drone is taking a U-Turn to change its lap.

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад +1

      According to me, this problem can be solved by these steps:
      You can pick detections only from the center of the image to reduce the chances of counting the same objects when the drone changes its perspective.
      1- Crop the input image to retain only the central region. The size of the central region can be adjusted based on your requirements.
      2- Perform object detection on the cropped central region of the image using YOLOv8 or your chosen detection model. This will focus the detection on the area of interest.
      3- Implement object tracking within the cropped central region. Assign unique IDs to detected objects and track them within this limited area.
      4- Count objects based on their unique IDs within the central region. When the drone changes its perspective or takes a U-Turn, you'll still be counting objects within the central region, which should help prevent double-counting.
      Keep in mind that the size of the central region you choose should be a balance between ensuring accurate object counting and maintaining a wide enough field of view to capture relevant objects. Experimentation and testing will be key to determining the optimal size for your specific application.

    • @pushpendrakushwaha604
      @pushpendrakushwaha604 Год назад

      @@CodeWithAarohi thank you

  • @palurikrishnaveni8344
    @palurikrishnaveni8344 Год назад

    One doubt shall I use yolov8 for skin lesion counting and detection purpose?

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      Yes, you can use. I recently did this video where I am detecting, tracking and counting using yolov8 : ruclips.net/video/Z3uquMElyzI/видео.html

  • @Computer-vision-Yolo
    @Computer-vision-Yolo Месяц назад

    There is any your repo ya video for person re-identification

    • @CodeWithAarohi
      @CodeWithAarohi  Месяц назад

      No, I don't think that I have posted any video on reid.

  • @Khelak007
    @Khelak007 Год назад

    I want only for "person" detection so what should i do some changes in this code??

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      just write classes = 0 where you are calling the model

  • @KiritJain-ww4zk
    @KiritJain-ww4zk 4 месяца назад

    Maam I am unable to find the "yolov8_DeepSORT" as when I clone the git repo it gives me the "deep_sort" folder but nothing else.
    Please help me if I am missing something as the code is unable to find the the images and giving me a "file not found" error

    • @KiritJain-ww4zk
      @KiritJain-ww4zk 4 месяца назад

      if anyone can help me with this problem

    • @CodeWithAarohi
      @CodeWithAarohi  4 месяца назад

      Email me at aarohisingla1987@gmail.com I will send you the code

  • @HIMANSHU-mt1jk
    @HIMANSHU-mt1jk 8 месяцев назад

    Maam can we make it real time

  • @narasimhang4736
    @narasimhang4736 5 месяцев назад

    does deepsort really need a dedicated gpu?

    • @CodeWithAarohi
      @CodeWithAarohi  5 месяцев назад

      No, You can work with cpu also but it will be slow as compare to gpu.

  • @saudshaikh5970
    @saudshaikh5970 Год назад

    i am working in google colab but cant get the output video
    i see that it is not releasing after the loop is done

    • @sumukharaghavanm6466
      @sumukharaghavanm6466 8 месяцев назад

      Facing the same issue. Found any solutions ?

    • @МаринаОрлова-ы2ч
      @МаринаОрлова-ы2ч 6 месяцев назад

      I also work at the collab. Did you manage to run the code? I'm having problems importing the DeepSORT method. How did you handle it?

  • @mamidipakasairoshini2305
    @mamidipakasairoshini2305 Год назад

    %cd yolov8_DeepSORT Mam where is this folder located?

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      github.com/AarohiSingla/Tracking-and-counting-Using-YOLOv8-and-DeepSORT

  • @mamidipakasairoshini2305
    @mamidipakasairoshini2305 Год назад

    Mam images/person.jpg is not present in github

    • @CodeWithAarohi
      @CodeWithAarohi  Год назад

      Create a folder with a name of " images" and then put image inside it.

  • @ganpatinatrajan5890
    @ganpatinatrajan5890 Год назад

    which software r u using to record ur videis

  • @welcomebro4218
    @welcomebro4218 10 месяцев назад

    video is created but there is no data :- code executed successfully and output video also created but it has 0kb data and can't open.