5. Dataset class for simple image / vision problems

Поделиться
HTML-код
  • Опубликовано: 26 дек 2024

Комментарии • 12

  • @rs9130
    @rs9130 3 года назад +1

    great work, where is the albumentations video?

  • @starlite5097
    @starlite5097 3 года назад

    Hi, thanks for the video. Great job explaining how this works.

  • @murugeshmanthiramoorthi2826
    @murugeshmanthiramoorthi2826 3 года назад

    It's awesome 🎉🎉

  • @sahhaf1234
    @sahhaf1234 2 года назад

    @4:35: Are you sure that augmentation is done at __getitem__?? To me it seems that augmentation must be done on the whole dataset before training, and it increases the number of images in dataset. Hence it should be a separate step and shouldnt be a part of __getitem__, whose task is to read a single item from the dataset..

  • @sahhaf1234
    @sahhaf1234 2 года назад

    thank you. but it is possible to augment an image in more than one way (reflection, playing with colors etc). which augmentation does this program do?

  • @manishsharma2211
    @manishsharma2211 3 года назад +1

    Quality 🔥🔥🔥

  • @akshay_pachaar
    @akshay_pachaar 3 года назад

    great stuff!!

  • @TheStealthPawn
    @TheStealthPawn 3 года назад

    Would it be better practice to do len(self.targets) since in a lot of cases, its just 1 image folder path and then you construct the exact image path for a given instance within __getitem__ itself? While targets is usually from a csv

  • @AlexeyMatushevsky
    @AlexeyMatushevsky 3 года назад

    Hi, thank you for your videos! I have a questoin about the `np.transpose function`. as i understand we transform the matrix of rows with pixeld into the separate trices of the separate channels per pixel... if it make sense.

  • @prudhvirajboddu6304
    @prudhvirajboddu6304 3 года назад

    If we have the images in respective folders for classification how can we pass targets and image paths to the dataset?

    • @abhishekkrthakur
      @abhishekkrthakur  3 года назад +1

      you can split on "/" to get the targets, then encode them using LabelEncoder of scikit-learn. everything else remains the same.

  • @AdityaJha1
    @AdityaJha1 3 года назад

    Hey Abhishek, can you suggest me how I should go about learning NLP in a structured way?, I already know the basics, but I wanted to go back to the drawing board and build a better foundation