TensorFlow Tutorial 18 - Custom Dataset for Images

[@ivanduranperez8115]

Keep making content like this, please. This has been extremely helpful. I hope your channel grows ❤️

[@UtkarshChaudhary-p4u]

Thanks a lot man. Was stuck on this for a long time. Literally a life saver.

[@ioannago3751]

You are amazing man ! I have seen the most of your videos and you cover so many subjects. You truly helped me a lot to figure out many questions that I had ! Great job , keep going !

[@kajalyadav4246]

This one is life saviour. Thank you. I have been going through documentations for hours & found nothing relatable. I was just about to give up for today that's exactly when I found you. You are amazing.

[@rajatvardam2787]

Exactly what I was looking for. And in multiple ways at that. This content has helped me. Please keep up the good work.

[@alierencelik388]

Just to say, I have been using PyTorch for 3 years and now I need to use TF. I have to say data loading in PyTorch is soo straightforward. In TF, you need to use TF's tools and prepare your folders accordingly. Thanks for the video.

[@HVossi92]

Awesome, I could not find any good explanation for the validation_split, thanks!

[@MaralSheikhzadeh]

Thank you. this really helped me create a custom dataset. :)

[@ilaydelrey3122]

this was immensely helpful!

[@AladdinPersson]

I was inspired and learned the basics of TensorFlow after I completed the TensorFlow specialization on coursera. Personally I think these videos I created give a similar understanding but if you wanna check it out you can. Below you'll find both affiliate and non-affiliate links, the pricing for you is the same but a small commission goes back to the channel if you buy it through the affiliate link which helps me create more future videos. affiliate: bit.ly/3JyvdVK non-affiliate: bit.ly/3qtrK39 Here's the outline for the video: 0:00 - Introduction 0:29 - Images in subfolders 14:03 - Folder with annotation csv 19:15 - All in single folder 24:53 - Ending For loading data that's structured inside of subfolders I also show you two different ways you can do it so that's why that one took a little bit longer than the others! :)

[@davidportilla4377]

thanks a lot! this is the tutorial i need! just new to deep learning , comunity so helpfull

[@vuledinh8262]

very helpful, thank you for this video

[@julienserbanescu1484]

I'm getting this error and i'm wondering what I can do to fix it: ' ValueError: Input 0 of layer dense is incompatible with the layer: expected axis -1 of input shape to have value 6272 but received input with shape [None, 20000]' I am using custom images but it is generally the same as the tutorial one, for the subfolder section. Any help would be appreciated thanks!

[@fingerstyledojo]

qBittorrent running in the background while using Mullvad VPN, exactly my setup! What a bro :3

[@sruthisriram9114]

In what format are the images of digits stored in your computer. jpg? png?

[@jamilahmarramduarte1575]

is it okay to use this in a many classes?

[@gadgetboyplaysmc]

Can you show how to resize the image for the "folder with annotation csv" example?

[@DizaFebriyan]

how to save the train model with this method ?

[@aimanjavid8441]

awsome Thank you so much

[@neemaradhalalneema5418]

Hi, google colab is showing keras is not defined in keras.sequential

[@sadeghkhosravi6154]

Thank you for this tutorial, when our model have multiple output, how to build database?

[@RC-qi6hs]

Good content, deserve a sub 🎉

[@divyankshah702]

Hey Aladdin plz help I am using method 1 for loading custom dataset stored in subfolders. Now, I want to use model.predict() on it but not able to use getting errors. Plz show how to use model.predict() in method 1

[@MultiAllan05]

Great tutorial, Aladdin! Do you know how to create custom TFDS datasets for object detection? It will help me a lot. Thanks!

[@afriquemodel2375]

i m lloking for this tutorial

[@eduardofdut]

Inside the function read_image, how can you resize the tensor?

[@rafaelzinnilopes8052]

In model fit, didn't you need a y_test and y_train to show that the number you have is 0,1,2,3,4,5,6,7,8 or 9?

[@stefano6551]

hi, if I have a folder with separated images in different folders (train test validation), how can i create a database? thank you

[@geordi-gabrielrenauddumoul449]

how to create a dataset that you can save and use to train on an other computer ?

[@Crimau12000]

Hey, thanks for the video. Cool explanation. Thought I have a question. How do handle regression problems? All of this tensorflow methods (flow from directory, Image generation) are build for classification probems.

[@Crimau12000]

Hey, thanks very much for the video. I have a question: When you build your dataset throuh maping images loaded with tf.io.readfile....., do you still need to build an image generator for feeding your model as you did in first part of the video?

[@jaydevisinghgohigo5646]

Very good jobdone

[@543resident4]

Hi Aladdin, I've been following your tutorials and they've been excellent so far. Have you ever come across the error 'ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape ()'? I got that error when I tried to load my own dataset with a csv file. Then, I tried downloading your exact code from github, but I still got the same error even when I ran that using the MNIST dataset and csv file that you used in this video. I'm pulling my hair out here, because there's absolutely no logical reason that code which you showed working in the video would suddenly not work on my own machine. If there's any help you could provide, it would be GREATLY appreciated.

[@patrickromero8724]

I have the same problem. Please I need help with this as well :(

[@mariuspy]

I have the same problem, I couldn't figure it out how to solve

[@ruvishkarathnayake9125]

Hi, can we use the same method to split the testing set as well or is this only for the ''training" and "validation" set splitting?

[@AladdinPersson]

No you can use it to split the test set as well!

[@robabmaghbol2247]

it was great example. where can we get the code and images.

[@emilefortier1688]

Hey Aladdin, I'm making my debut into GANs and was wondering if your guide here would translate well for GAN training. Thanks for the awesome content!

[@_ABDULGHANI]

thank you for this tutorial, I have a dataset with XML annotation and all the objects inside one folder (person, car, bus, motorcycle) I want to convert this dataset to CSV or any format accepted by PyTorch or Tensorflow do you have any idea how can I convert them? I want to use this dataset to train with faster R-CNN algorithms with Resnet.

[@starstenaal527]

is validation the same as testing?

[@rodizadi9591]

No, Validation is used to tweak weights as your model is training. Testing data is used to evaluate your model after the training process is complete. You wouldnt want to use validation data for testing because it your model has seen those images before which will play into biases

[@RajatSharma-ct6ie]

Hi Aladdin, thanks for the wonderful tutorials. I have another requirement in which in have multiple inputs: [ds_1, ds_2] and output as y1 , so ds_train = ([ds_1, ds_2], y1). Also, the augmentation is applied only on ds_1. Where the count of ds_1 & ds_2 is equal. How to pass it in Model.fit ()? Any references or suggestions?

[@Cyco2k5]

Hey, did you ever figure out how to do this? Its is exactly what im trying to do too

[@seyeeet8063]

thanks for the video, It would be very useful if you can covering other important things that are useful for dataloader, such as how to do custom augmentation in the dataloader, how to use stuff like tfrecord, etc that are designed for dataloader and what is theri advantages, how to do dataloader if we have very large dataset (e.g. more than 100K images) thanks

[@radualexa3713]

Hello! It's a very good video...Thank you! Could you tell me how can I upload my own images to predict them with the model I trained, please? I try to find this information an until now I could not find anything about this subject.

[@underdog5255]

My model keep predicting same class everytime and it's not even a class from my datas3t

[@ranjithganesan736]

How to convert tflite ?

[@hamidesmaeilifardi800]

Good

[@meliskara-n6x]

Are your images grey?

[@ambujmittal6824]

@23:46 In some way? What is that way? 😅

[@universe3406]

Wow, Another video. I am working on it right now. I need some help, my dataset has 5 classes. It detects them perfectly. I have used pre-trained model SSD mobile net_V1. Now the model detects other objects too. I tried removing the fine_tune_checkpoint, Yet it detects other objects like a laptop, Painting, Sometimes, humans too. How can I overcome it?

[@AladdinPersson]

I'm not really sure sorry, I don't have much experience in object detection, actually researching more about it right now

[@bellemarravelo6002]

NameError Traceback (most recent call last) in 12 13 color_mode='grayscale', ---> 14 batch_size=batch_size, 15 image_size=(img_height, img_width), 16 shuffle=True, NameError: name 'batch_size' is not defined

[@bloodraven9622]

Error : logits and labels must have the same first dimension

[@adamlakatos7987]

Hero

[@krishnapraneeth5448]

I need to contact you

[@rozanwali8345]

Ihjhh