Tensorflow Input Pipeline | tf Dataset | Deep Learning Tutorial 44 (Tensorflow, Keras & Python)

Поділитися
Вставка
  • Опубліковано 18 гру 2024

КОМЕНТАРІ • 85

  • @codebasics
    @codebasics  2 роки тому +3

    Check our Deep Learning Course (in PyTorch) with Latest, Industry Relevant Content: tinyurl.com/4p9vmmds
    Check out our premium machine learning course with 2 Industry projects: codebasics.io/courses/machine-learning-for-data-science-beginners-to-advanced

  • @pravallikaece195
    @pravallikaece195 Рік тому +24

    You are helping the data science community in an excellent way. keep going on and all the power to you. Thanks! and a very small token of appreciation

  • @pintokatendejonathan1740
    @pintokatendejonathan1740 3 роки тому +6

    Thank you expert, I follow you from the democratic republic of the congo and I appreciate everything you do

  • @anmolkhurana490
    @anmolkhurana490 2 місяці тому

    I was stuck with this Input Pipeline code for my project since last week. but, you cleared my all problems in just one video. Hats off to you for explaining such complex concepts in the easy way 👏

  • @muhammadmalik4627
    @muhammadmalik4627 2 роки тому +1

    Very useful tutorial! This video shows the Tensorflow Process in a simple and sistematic way. And your explanation is far clearer than any other expert tutorial. Big thumbs up for you, Sir!

  • @dec13666
    @dec13666 3 роки тому +15

    Holy smokes, TF pipeline looks so easy now! What a nice tutorial! 😎👍 TF pipelines always looked like black magic to me... until now! 😂... Keep it up with the good work, pal!

  • @ahmedyaseen8994
    @ahmedyaseen8994 Рік тому

    i love you man. Been struggling with tf for 2 months as I only have experience with pandas. The theory part was so helpful in understanding why tf is the way it is. And obv the coding part too. Thank you so much!

  • @lebesgue-integral
    @lebesgue-integral 3 роки тому +1

    Changing ImageDataLoader to Dataset boosted my training time :) Thank you

  • @shantanubapat6937
    @shantanubapat6937 3 роки тому

    Only 175 likes? This video should have like 100k likes. Good content!

  • @fernandotta
    @fernandotta 2 роки тому

    Man, this really helped me out. I was overcomplicating things. Thanks a bunch!

  • @freedmoresidume
    @freedmoresidume 2 роки тому +1

    Awesome tutorial, you never disappoint 😎👍

  • @someone7421
    @someone7421 3 роки тому +5

    I have seen many of your videos and all are so informative. You should make reinforcement learning tutorials as well and best of luck for your future videos

  • @akshaypatwal6013
    @akshaypatwal6013 3 роки тому +2

    Great work sir , learnt a lot from ur videos and looking forward to it in future also

  • @Noir_SD
    @Noir_SD 3 роки тому

    It was really helpful. Thank you so much for this awesome tutorial on tensorflow data pipeline. Keep making this type of videos more.😀😀😀😀

  • @sanjuvikasini1598
    @sanjuvikasini1598 4 місяці тому

    Amazing explanation sir!

  • @shivkrishnajaiswal8394
    @shivkrishnajaiswal8394 2 роки тому +3

    is DataSet API useful for small dataset (data enough to be in RAM). It seems in every EPOCH of training, all the files will be reloaded. Which could be slow.

  • @shubhamtalks9718
    @shubhamtalks9718 3 роки тому

    Jaan bacha li guruji🙌

  • @aruneshprasad3064
    @aruneshprasad3064 3 роки тому +1

    You're excellent sir😇

  • @nasgaroth1
    @nasgaroth1 3 роки тому +4

    Awesome, but will be better if you could show how to uses it with tensorflow model, what is not such straitforward like it looks

  • @kevian182
    @kevian182 Рік тому

    Excellent tutorial! Thank you

  • @AyadEKorial
    @AyadEKorial 2 роки тому

    Great instructor 👍🏻😎

  • @favourogbenna4402
    @favourogbenna4402 Місяць тому

    Superb tutorials

  • @inesriahi
    @inesriahi 3 роки тому

    Extremely useful !
    Keep going !

  • @jorismattheijssens5238
    @jorismattheijssens5238 2 роки тому +1

    Great video! The slides are neat, the explanations are clear and to-the-point. One question: I want to figure out how to stop the shuffling of a tf.data.Dataset every time you use a function, but I couldn't figure it out yet. For example, at 25:39, you extract the labels, but they are not the same as those in the file paths on the cell above. Any idea how NOT to shuffle the instances in a dataset?

  • @VerdadesOlvidadas
    @VerdadesOlvidadas 3 роки тому

    Nice video and humor thanks.

  • @jacksonngari1670
    @jacksonngari1670 Рік тому

    i wish to learn on both deep learning and python through you.

    • @codebasics
      @codebasics  Рік тому

      I already have a python course on codebasics.io. For data science, ML and deep learning I will be launched courses on codebasics.io in the coming few months. Stay tuned 🔥

  • @biggusmaximus1651
    @biggusmaximus1651 3 роки тому

    Very great video sir

  • @niranjansitapure4740
    @niranjansitapure4740 Рік тому

    This was crazy useful!

  • @pvtgcn8752
    @pvtgcn8752 2 роки тому

    Great Explanation

  • @MdAbdullah1509
    @MdAbdullah1509 3 роки тому

    30:46 after 6 iteration it shows error while printing rest, how to fix this error
    Input/output error
    [[{{node ReadFile}}]] [Op:IteratorGetNext]

    • @TayyabaSajjad-j3f
      @TayyabaSajjad-j3f 5 місяців тому

      I am facing same error in colab. How to fix it? Anybody's help is appreciated😢

  • @mvs69
    @mvs69 3 роки тому

    can anyone explain 25:39 how get_lable method receives file path while calling below function,
    for label in train_ds.map(get_label):
    print(label)

  • @riyadislam5412
    @riyadislam5412 3 роки тому

    Great work sir,,We need tutorial fuzzy based ML and DL.

  • @kushalgandhi3520
    @kushalgandhi3520 2 дні тому

    What if, it is numpy array file that is required to be read parallel to training?

  • @devloper_hs
    @devloper_hs 3 роки тому

    at time 10:03 the code will be
    for sales in tf_dataset.as_numpy_iterator():
    print(sales)
    and not
    for sales in tf_dataset.as_numpy_iterator:
    print(sales)
    may be some changes in new version

  • @haneulkim4902
    @haneulkim4902 Рік тому

    Thanks for great explanation! I've got two questions.
    1. You said that it loads data in batches from disk how does shuffling work? Data are sampled from multiple source data then made into one batch or somehow all data is shuffled from disk?
    2. I am trying to write tfrecords from pandas dataframe, how to split x,y within tf.data.dataset so it can be trained? After reading tfrecords I have dictionary of features(tensors).

  • @amirayasmine8490
    @amirayasmine8490 3 роки тому

    great explanation .. thank you

  • @_Ahmed_O
    @_Ahmed_O 10 місяців тому

    Awesome ! Thanks a lot.

  • @mubashirayub6630
    @mubashirayub6630 10 місяців тому

    My dataset files are in .npy format, I want to fetch these files as you did for images by using image.decode_jpeg() fucntion. I couldn't find any function to fetch data from .npy file in Tensor.
    Your response would be appreciated...

  • @dheemanth_bhat
    @dheemanth_bhat Рік тому +2

    if anyone gets this error: `InvalidArgumentError: Unknown image file format. One of JPEG, PNG, GIF, BMP required.`
    just delete file `Best Dog & Puppy Health Insurance Plans....jpg` in dogs folder.

  • @frankieiero6859
    @frankieiero6859 Рік тому

    does this input pipeline also applicable for hyperspectral images?

  • @harshalbhoir8986
    @harshalbhoir8986 Рік тому

    This is awesome!!!!

  • @siddyant8332
    @siddyant8332 2 роки тому +1

    Sir could you please cover neural structured learning package, specifically the adversarial regularisation and graph regularisation topics from it, since there aren't many videos on youtube regarding these ....

  • @lequoctuan8848
    @lequoctuan8848 2 роки тому +1

    Hello sir, this video is very helpful, thank you for creating this. My question is, when I use model.fit after building an input pipeline for training set, should I use validation_split for each batch for validation or should I use dataset.skip() to create validation set and then use it to validate for every training batch? Sorry for bad grammar!

  • @kutilkol
    @kutilkol 3 роки тому

    so how to create the batches online during training and how to pass the batches to the model during training. not shown in video

  • @user-qz4lw3cs8e
    @user-qz4lw3cs8e 2 роки тому

    If I get some image matrix data and save it as a dataframe, how do I pass it into the dataset as a feature? The from_tensor_slices method will report "ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray)." Thanks everyone for your help!

  • @devloper_hs
    @devloper_hs 3 роки тому

    train_ds.map(process_imgs) returns : TypeError: tf__process_imgs() takes 1 positional argument but 2 were given , how to fix

  • @kolatimisimi7879
    @kolatimisimi7879 2 роки тому

    Nice Tutorial, thanks.
    Also I think you could have just included the scaling part in the process function

  • @akrsrivastava
    @akrsrivastava 3 роки тому

    Good introductory tutorial.
    Why is there a b' in front of the file paths? The b' usually signifies byte data, doesnt it? Then how come it allowed to do a string split() ?

  • @peterpirog5004
    @peterpirog5004 2 роки тому

    Very nice tutorial. I wonder how to generate dataset with random numbers, for example vector with uniform distribution in range with defined size to use while fiting with defined number of epochs and defined batch size. Is possible to use for this purpose tf.data.experimental.RandomDataset in tf 2.10 ?

  • @srinathblaze651
    @srinathblaze651 Рік тому

    What if folders are not clearly separated as cats and dogs.. and we have just one folder of all images of cats and dogs.

  • @sergiochavezlazo5362
    @sergiochavezlazo5362 Рік тому

    What if instead of creating a new function scale, you just add one more line to the previous function:
    img=img/255 #Normalize

  • @aravindetagi4008
    @aravindetagi4008 2 роки тому +1

    I am watching this video after facing this issue.... 12GB RAM of google colab got filled and runtime crashed for loading 16k images.... Then I started using ImageDataGenerator BatchDataset

  • @lahiru954
    @lahiru954 2 роки тому

    It's awesome and thank you. But I want to ask a question. How can we apply the same concept for video data (already framed). can someone explain please

  • @parthsharma898
    @parthsharma898 2 роки тому

    Sir can you please explain that how can we convert rgb images into array? @codebasics

  • @nitinlodha4561
    @nitinlodha4561 3 роки тому

    image_count=len(images)
    print(image_count)
    TypeError: object of type 'TensorSliceDataset' has no len()
    Am getting this error , how to solve it?

  • @sakethgupta9047
    @sakethgupta9047 2 роки тому

    Images\*\* how to give my file input in that place?
    I also stored cat and dog pictures in same Library

  • @alanhenry9850
    @alanhenry9850 3 роки тому

    Sir can u also upload videos on bigadata,kafka,spark

  • @enewhuis
    @enewhuis 2 роки тому

    Enjayable presentation. But I have 64GB on MY laptop. :P

  • @snehasneha9290
    @snehasneha9290 2 роки тому

    Hello, sir please suggest to me some projects for my masters. now I am studying MSC on data science I want to do a data science master project

  • @manideep1882
    @manideep1882 3 роки тому

    Videos are getting a little blurry, other than that it was a very informative. I've tried the shuffle and map combination and TF makes life easy. TY

  • @niranjandattatreya8331
    @niranjandattatreya8331 2 роки тому

    Hii, thanks a lot for the video , very useful, can you please upload tutorial on creating a custom dataset from parallel corpus of data for training ? unable to figure out

  • @vijayakakumani613
    @vijayakakumani613 3 роки тому

    very nice tutorial on Tf.data.Dataset module! my question is, if we use ImageDataGenerators will all this be automatically done? I.e. both creating Image Input Pipeline and also optimizing the pipeline (which is covered in next tutorial)

  • @Amir-gi5fn
    @Amir-gi5fn 5 місяців тому

    I saved my X_train to a binary file how load it as tensor to make it batches

  • @swatikarot8272
    @swatikarot8272 2 роки тому

    tf_dataset = tf.data.Dataset.list_files('.\\datasets\\flower_photos\\*\\*', shuffle = False).map(lambda x: process_image(x)).map(lambda x,y: scale(x,y))
    Could someone review this one line code for image dataset??

  • @waadturki2359
    @waadturki2359 10 місяців тому

    I want to process a video data set anyone has a hint or a similar YT video

  • @shwetameena0511
    @shwetameena0511 Рік тому

    from 20:05

  • @OmidAtaollahi
    @OmidAtaollahi 3 роки тому

    perfect, tnx.

  • @neerajashish7042
    @neerajashish7042 2 роки тому

    If any one facing this error, where "tensorflow object has no len()" , instead of len(image_ds) use len(list(image_ds))

  • @aditya_01
    @aditya_01 3 роки тому

    thank u sir

  • @peopleclickindia1733
    @peopleclickindia1733 3 роки тому

    The Python Training is designed for you to learn Python Programming Principles such as Python Program Flow, Control Statement, Loops, Functions, File Handling, Error Handling, and Access to APIs. Feasible lessons of real-time tasks and test cases are the subject of this Python course in BTM Layout. We are Bangalore's leading educational institution with Job-Oriented Python Training in HSR Layout. Being the best Python training institute in Bangalore, We understand the latest trends and provide the best Python training in Bangalore. Learn with brilliant teachers and gain outstanding programming skills in Python. Turn your fears into opportunity and get your dream job.
    If you have the commitment and perseverance to make your career dreams true, connect with us without any second thoughts. We are here to help you with our hyperactive team and full-fledged learning facilities to nurture your skills and polish them to match up to industry standards. Here are the prime features that make us unique.
    * 100% placement guaranteed training
    * Resume building support
    * Training by Industry experts
    * 24/7 solution support
    and many more
    So you can start your career as a Python programmer with our python training in Bangalore. Contact us now.
    Phone no: +91 76193 43002
    Website: www.peopleclick.in/python-training-bangalore.html?SEO&Python

  • @navneet3211
    @navneet3211 3 роки тому

    ❤️

  • @samas69420
    @samas69420 2 роки тому

    SSSSSINGGGLL LINE OFKODE

  • @tech-learner4555
    @tech-learner4555 5 місяців тому

    Are you Indian?

  • @shantib4025
    @shantib4025 Рік тому

    tf_dataset = tf_dataset.filter(lambda x: x>0)
    for sales in tf_dataset.np():
    print(sales)
    AttributeError Traceback (most recent call last)
    in
    1 tf_dataset = tf_dataset.filter(lambda x: x>0)
    ----> 2 for sales in tf_dataset.np():
    3 print(sales)
    AttributeError: 'FilterDataset' object has no attribute 'np'

  • @adrenochromeaddict4232
    @adrenochromeaddict4232 9 місяців тому

    what you promise to show fixing and what you actually show have nothing to do with eachother. and it's so emberrassing that as if botting your sub count wasn't enough you're botting your comments section too. another pajeet wasting my time