Object Tracking using YOLOv8 on Custom Dataset

Поділитися
Вставка
  • Опубліковано 18 лис 2024

КОМЕНТАРІ • 91

  • @arnavthakur5409
    @arnavthakur5409 Рік тому +1

    Thank you so much for sharing this video. It helped me a Lot. Kindly keep sharing ma`am

  • @OmidFatemi-y6z
    @OmidFatemi-y6z 22 дні тому

    thanks for the video , very educative, may i ask, where is the git for feature extractor custom training?

  • @SP-db6sh
    @SP-db6sh Рік тому

    Eagerly waiting for your video on using this yolov8 micel in drone to detect object real-time & send notification to db or other Kafka consumer !

  • @user-maomao-tsai
    @user-maomao-tsai Рік тому

    Thank you for your great video @ AArohi

  • @vipulupadhyay8726
    @vipulupadhyay8726 8 місяців тому +1

    I am using yolo v8 , with bytetrck , some time track Id change,could you please share configuration of bytetrack

  • @Sunil-ez1hx
    @Sunil-ez1hx Рік тому

    What an amazing video.. Worth watching. It again & again

  • @kanikabisht6331
    @kanikabisht6331 Рік тому

    Best video of VOLOv8

  • @MIGUEL-bv1qh
    @MIGUEL-bv1qh Рік тому

    Well done!!. It’d be nice to incorporate counting for one class and multiple ones and saving results in csv, this is one of the challenge thing in CV

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому +3

      Thankyou for the suggestion. Will try to cover the suggestion in upcoming videos 🙂

  • @soravsingla6574
    @soravsingla6574 Рік тому

    VEry nicely explained

  • @afrinashraf3914
    @afrinashraf3914 8 місяців тому

    Is the data annotated seperately using annotation tools since it seems boat_yacht_dataset folder and dataset folder with train and valid(with labels and images) and simply train and val folders respectively.

  • @arnavthakur5409
    @arnavthakur5409 8 місяців тому

    Well done

  • @shinwarikhan4677
    @shinwarikhan4677 Рік тому

    Thank you so much mam! i have been waiting for this video for a long time

  • @saiakshithaa
    @saiakshithaa 8 місяців тому

    I am getting unknown model error it mentions the model must be from resnet18,resnet34,resnet50…etc I have built my own weights for yolo weights and Reid weights kindly help me to solve this

  • @FREEFIREGAMER-iv8dx
    @FREEFIREGAMER-iv8dx Рік тому

    Madam pls explain semi supervised learning techniques on yolov8

  • @madushangunarathne353
    @madushangunarathne353 Рік тому +1

    Madam how to do the feature extraction of a dataset which has multiple classes.(Like Cats & Dogs). Here only having a single class which is boat class.

    • @sarahch8878
      @sarahch8878 8 місяців тому

      Hello, did you have an answer to your question?

  • @aayushgupta1034
    @aayushgupta1034 Рік тому +1

    I am training object tracking model on visdrone dataset. I followed exactly same approach as mentioned in video. But while performing visualization in "visualize_model" function, it is giving error: `Invalid shape (3, 224, 224) for image data` in line: `plt.imshow(inputs.cpu().data[j])`. also the model so trained is getting rejected while performing the actual tracking..(saying it is not of resnet18 type).. Please let me know where I may be wrong..Thank you mam for providing a good resource in domain of object tracking.

    • @aayushgupta1034
      @aayushgupta1034 Рік тому +1

      @CodeWithAarohi I think you need to use plt.imshow instead of only imshow since it will give error. Kindly do correct me if I'm wrong.

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      If plt.imshow is working then use it. Otherwise as per your error: the error message indicates that the shape of the input tensor is (3, 224, 224). The first dimension (3) likely corresponds to the number of color channels in the image (e.g., red, green, blue), while the next two dimensions (224, 224) correspond to the height and width of the image.
      To fix this error, you may need to reshape your input data to match the expected shape of the plt.imshow() function. One way to do this is to use the transpose() function in PyTorch to change the order of the dimensions in your tensor: # Assume that "inputs" is your input tensor with shape (3, 224, 224)
      inputs = inputs.transpose(0, 2).transpose(0, 1) # Transpose the dimensions to (224, 224, 3)
      plt.imshow(inputs.cpu().data[j]) # Display the image

    • @aayushgupta1034
      @aayushgupta1034 Рік тому

      @@CodeWithAarohi I use plt.imshow(inputs.cpu().data[j].T) # Display the image
      to make dimensions as (224, 224, 3) but now in the output the images are like film negatives which are not relevant to val dataset

    • @aayushgupta1034
      @aayushgupta1034 Рік тому

      @Code With Aarohi, please share link of this notebook. I will cross check the code. Maybe I am wrong somewhere. Thank you!!

  • @Anupkumar-cd6gv
    @Anupkumar-cd6gv Рік тому

    Hello arohi, how we can use custom model.weight in production as i am working with yolov5 do we need clone whole official repo ?

  • @mahejabeenhossain9727
    @mahejabeenhossain9727 6 місяців тому

    Thank you for this great video on how to prepare a dataset for feature extractor model! Is this the same way for ReID?

  • @sarahch8878
    @sarahch8878 8 місяців тому

    Thank you very much for these explanations, How can we organize our dataset for feature extration models if we have two classes for example?

  • @waseemhaider5295
    @waseemhaider5295 9 місяців тому

    mam can you apply this code on person detection and tracking

  • @hendragunawan119
    @hendragunawan119 Рік тому

    terima kasih ilmunya

  • @ojasmittal
    @ojasmittal Рік тому

    Hello Maam, I am trying to use Yolo V8 for object detection through camera and along with that i want coordinates of the bounding boxes. Can you please guide me how can i do that?

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      You can extract the coordinates of the bounding boxes by parsing the output of the YOLOv8 model.

  • @AshokanandChaudhary786
    @AshokanandChaudhary786 Рік тому

    How can we use the yolo in the extremly large dataset i.e (like having 1 lakh images with more objects) ? Wouldnt it will be difficult to manually anotate the larger set..?

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому +1

      Yes, It will be a time consuming process but you have to annotate the whole dataset. There are organizations who provide paid annotation services.

  • @shinwarikhan4677
    @shinwarikhan4677 Рік тому +1

    mam did you upload it to your github repo

  • @shinwarikhan4677
    @shinwarikhan4677 Рік тому

    thank you so much

  • @SimranSingh-cp2el
    @SimranSingh-cp2el Рік тому

    hi! is your jupyter notebook available somewhere?

  • @SaimKhalid-wx2zv
    @SaimKhalid-wx2zv Рік тому

    Thank You for sharing this, can we draw tracks too using this code?

  • @MotivationBabyBoss
    @MotivationBabyBoss 7 місяців тому

    can we also use that feature extractor to recognize humans individually?

    • @CodeWithAarohi
      @CodeWithAarohi  7 місяців тому +1

      Yes, feature extractors can indeed be used for recognizing humans individually

    • @MotivationBabyBoss
      @MotivationBabyBoss 7 місяців тому

      @@CodeWithAarohi is this folders and code available in github by chance. bc its very important for my FYP research.🙂😊

    • @MotivationBabyBoss
      @MotivationBabyBoss 7 місяців тому

      @@CodeWithAarohi is that code and folders available on GitHub by chance ?this project is very important for my FYP researches. thank you very much

  • @anitanbhatt
    @anitanbhatt Рік тому

    can i use json file as input?

  • @_Fangirling
    @_Fangirling Рік тому

    Can you please tell me how to calculate the overall accuracy(SHOWING mAP score) of the particular trained yolov7 model in Google colab?

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому +1

      from utils.metrics import ap_per_class
      # define the confidence threshold and IoU threshold for non-maximum suppression
      conf_thres = 0.25
      iou_thres = 0.45
      # define the test set of images
      test_images = ['path/to/your/test/images']
      # define the number of classes in your dataset
      num_classes = 1
      # generate predictions on the test set
      prediction = []
      targets = []
      for path, img, im0s, _ in LoadImages(test_images, img_size=640):
      img = torch.from_numpy(img).to(model.device).float()
      img /= 255.0
      if img.ndimension() == 3:
      img = img.unsqueeze(0)
      with torch.no_grad():
      prediction += model(img)[0]
      targets += LoadImages.detect(exists=True)[0][1]
      # calculate the AP per class
      metrics_output = ap_per_class(prediction, targets, plot=False, fname="precision-recall_curve.png", imgsz=imgsz)
      mAP = metrics_output['mAP'][0]

  • @soravsingla8782
    @soravsingla8782 Рік тому

    Thank you so much mam

  • @nnaufalir
    @nnaufalir 6 місяців тому

    what is difference obect detection and object tracking

    • @CodeWithAarohi
      @CodeWithAarohi  6 місяців тому

      Detection involves placing a bounding box around an object and tracking means determining whether that object appears in next frames and for how long, by assigning a unique ID to each object.

  • @techinfo89
    @techinfo89 Рік тому

    I have a problem where it says graph execution error..could you make a video on this?

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      Pleas explain the error in more detailed way. I will try to help

    • @techinfo89
      @techinfo89 Рік тому

      @@CodeWithAarohi mam I have a football dataset and I'm predicting the winner for a betting bot,I'm using the LSTM model ...on the first epoch it gives the error says "graph execution error"

  • @nit8826
    @nit8826 Рік тому

    I have a doubt I trained my model but how to count number of object of that label

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      Will try to cover in upcoming videos

    • @nit8826
      @nit8826 Рік тому

      @@CodeWithAarohi please just try to explain or make as soon as possible it's urgent project 😅

  • @Rehantv-g2s
    @Rehantv-g2s 10 місяців тому

    best_model_wts = copy.deepcopy(model.state_dict())---- this line takes infinite time to execute, pl help to reduce the time. Is it because of HD quality of images?

  • @hamidraza1584
    @hamidraza1584 8 місяців тому

    Please share the link of this feature extractor on pytorch official website

    • @CodeWithAarohi
      @CodeWithAarohi  8 місяців тому

      pytorch.org/vision/stable/feature_extraction.html

  • @gbo10001
    @gbo10001 Рік тому

    i couldn't find the guide in pytorch docs. can you share the code?

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      pytorch.org/tutorials/beginner/introyt/modelsyt_tutorial.html

  • @anushathatikonda3855
    @anushathatikonda3855 Рік тому

    Hi Ma'am, Can you please share the code

  • @yoonnouhan8057
    @yoonnouhan8057 10 місяців тому

    How to use custom dataset on real-time project? Can anyone guide me

    • @CodeWithAarohi
      @CodeWithAarohi  10 місяців тому

      The process I have explained in this video is for custom dataset.

  • @kuldeepaher4937
    @kuldeepaher4937 Рік тому

    Can you please provide collab liink?

  • @akbabu3007
    @akbabu3007 Рік тому

    Mam i need the dataset from this video can i get it from u please ..please reply mam

  • @aqibrehmanpirzada4552
    @aqibrehmanpirzada4552 Рік тому

    Where is the Source Code or Git Repo Link Ma'am ?

  • @ishankamadhuranga3729
    @ishankamadhuranga3729 Рік тому

    How predict movement of object

    • @CodeWithAarohi
      @CodeWithAarohi  Рік тому

      Please elaborate your question

    • @ishankamadhuranga3729
      @ishankamadhuranga3729 Рік тому

      @@CodeWithAarohi i want to identify object for example man then i want to predict that man go to right or left

  • @sarthaksamantaray6565
    @sarthaksamantaray6565 Рік тому

    Dataset kahan hai ma'am. plz provide.

  • @meghakataria8571
    @meghakataria8571 Рік тому

    can you please share the code

  • @biggiewalata
    @biggiewalata Рік тому

    Thanks!

  • @saiakshithaa
    @saiakshithaa 8 місяців тому

    Can you share your folder link please🥹🥹

    • @CodeWithAarohi
      @CodeWithAarohi  8 місяців тому

      This code is available for channel members (Contribution level 2)

  • @charankumaruttaravalli4910
    @charankumaruttaravalli4910 2 місяці тому +1

    Mam i need the dataset from this video can i get it from u please ..please reply mam

    • @CodeWithAarohi
      @CodeWithAarohi  Місяць тому

      This is available for channel members (Contribution level 2)