thank you, that's so great tutorial. hope thah you can continue to make tutorial about detection and other using deep learning. thank you so much 🙂
@atiqahzulkifli8556
6 ай бұрын
hi i want ask why error at model.fit? can you help me
@SOMEONE-ik3gp
2 жыл бұрын
Great Work.
@DeltonFerns
2 жыл бұрын
I'm waiting for part 2 😜
@IJApps
2 жыл бұрын
Glad to hear that. If you haven't already, subscribe to get notified when it releases!
@KEVINPRATAMASINAGA
Жыл бұрын
Hi, I want to ask, can this image classification app detect multiple objects in the same image?, btw sorry if my english is bad
@juanandresveravera7952
2 жыл бұрын
can you explain how use tensorboard in this project pls :(
@angelikakreutzer3182
Жыл бұрын
Please help me import numpy as np model.fit( train_ds, validation_data=val_ds, epochs = 10 ) I get an Graph Execution error
@angelikakreutzer3182
2 жыл бұрын
How you can integrate zip file into bitbucket? Whats possible there is to writing a txt.file. please help me out or better, make a Video. Thank you
@IJApps
2 жыл бұрын
Regarding bitbucket, the only thing you need to do is run this command: !wget bitbucket.org/ishaanjav/code-and-deploy-custom-tensorflow-lite-model/raw/a4febbfee178324b2083e322cdead7465d6fdf95/fruits.zip
@didin1290
2 жыл бұрын
Hi, what if my image in my dataset is not in a 1:1 ratio? I tried using my non-1:1 image and it stretched out after plotting them. Will this stretched image affect the prediction when the model is deployed in Android Studio?
@IJApps
2 жыл бұрын
If the image has been drastically stretched, it might affect the prediction.
@DeltonFerns
2 жыл бұрын
Great video!
@notfromthisworld2127
2 жыл бұрын
Hi. Does this dataset contain a dedicated folder or a file for all class labels?
@GeorgeTrialonis
Жыл бұрын
Bravo! You are great in explaining ML concepts and ideas. Very instructive tutorial. Thank you!
@akhil.o_s
2 жыл бұрын
how to use mobileNet in this project
@khamismuniru5188
11 ай бұрын
Hello, I have a question, I am working with many classes, how do I create my list of class names? "class_names = ["apple", "banana", "orange"]", I have tried to use "mylabels = os.listdir('/content/images/train')" but this just seems to give a random order list.
@aureocosta4994
2 жыл бұрын
It's so good! thanks for sharing the knowledge.
@momnaazmat371
2 жыл бұрын
How can we create link of zip file of our image data set in bitbucket?
@IJApps
2 жыл бұрын
You don't need to create your own link. You can check the video description for the command. This is the link to the bitbucket zip file: kzitem.info?event=video_description&redir_token=QUFFLUhqazRHd0lCTnQzT3dUcXRXOFBhX0xibWZockMyd3xBQ3Jtc0tudDJoODFrd0FBaDJFc1dmUGJKUG1WTExtNHlvSWpOM0Q0d0JiQnVpbFNQVG5GcDJXUzl1SFB0Tk51VHJiVlNXRWkzQlRaUHFHeGw5RTJreFBiUi1URkhvT1lWU0NnTDRDaEFiTk93S0x4em5EYnRiMA&q=https%3A%2F%2Fbitbucket.org%2Fishaanjav%2Fcode-and-deploy-custom-tensorflow-lite-model%2Fraw%2Fa4febbfee178324b2083e322cdead7465d6fdf95%2Ffruits.zip&v=ba42uYJd8nc&html_redirect=1
@subramanyabellary3939
2 жыл бұрын
which architecture this model is using?
@Liev04
2 жыл бұрын
Amazing work, god bless your kindness
@MiguelNFer
2 жыл бұрын
This is amazing :) great tutorial, keep going
@sirnikzchannel7107
2 жыл бұрын
Great video! I'm waiting for part 2,Thank you so much!
@KonstantinosKalogirou
Жыл бұрын
A wonderful, clear and fully understandable tutorial
@ericoelloco6515
Жыл бұрын
Hi, very nice tutorial but I have a question. Why array of confidences consist of some random numbers for me and we decide based on the greatest value. I mean, why that numbers are not in the range 0-1 as in the model from teachable because from that model I can't display confidence of my predictions.
@issanasralli4529
Жыл бұрын
Well done!!! Very good video!!! did you make a video about object detection using android studio? (Classification+localisation)
@alostsoul9594
Жыл бұрын
Done all steps but when trying to import shows Please choose tflite model though I chose one
@fawniaavissa9711
2 жыл бұрын
so helpful! im waiting for part 2! so excited
@neosebas8272
2 жыл бұрын
Gracias
@vil9386
Жыл бұрын
Awesome tutorial. Very easy to follow.
@asmaghanty7597
2 жыл бұрын
Hi, have you created your own dataset?
@IJApps
2 жыл бұрын
No. I used a dataset on Kaggle and reorganized it for this tutorial. You can download it at this link: kzitem.info?event=video_description&redir_token=QUFFLUhqazRHd0lCTnQzT3dUcXRXOFBhX0xibWZockMyd3xBQ3Jtc0tudDJoODFrd0FBaDJFc1dmUGJKUG1WTExtNHlvSWpOM0Q0d0JiQnVpbFNQVG5GcDJXUzl1SFB0Tk51VHJiVlNXRWkzQlRaUHFHeGw5RTJreFBiUi1URkhvT1lWU0NnTDRDaEFiTk93S0x4em5EYnRiMA&q=https%3A%2F%2Fbitbucket.org%2Fishaanjav%2Fcode-and-deploy-custom-tensorflow-lite-model%2Fraw%2Fa4febbfee178324b2083e322cdead7465d6fdf95%2Ffruits.zip&v=ba42uYJd8nc&html_redirect=1
@johncedricpamigao7811
Жыл бұрын
@@IJApps it's allowed to create my own dataset?
@lilymolejon5636
2 жыл бұрын
What if you take a picture of any other fruit what will happen?
@IJApps
2 жыл бұрын
Right now, the model as we trained it only outputs whether an image is an apple, orange, or banana. So if we give it a picture of a strawberry, it will still try to classify it as an apple, orange, or banana. However, you could use the model's confidence in its classification to display "unknown" if the confidence is really low.
@lilymolejon5636
2 жыл бұрын
@@IJApps Yea i have the same problem, I've been looking for a possible solution to that problem. Do you have any recommendation that when you take a picture of the other friuts "Unknown Fruit" will appear in Classification and will automatically back at the camera
@IJApps
2 жыл бұрын
@@lilymolejon5636 The question you ask is a really good one. I do no tknow how to have the machine learning model actually provide "Unknown fruit" as a class besides training it on a bunch of random images, but even then there are simply too many possible non-orange/banana/apple objects out there. Besides using our current model's confidences, one idea I have is using a type of model called an Autoencoder. Basically an autoencoder is a model that learns to reconstruct images. So if you train an autoencoder on orange, banana, and apple images, it will try to output that same orange, banana, apple image. The reason this is helpful is because now, if your trained model is given a coconut image, since it has only learned what apples, oranges, and bananas look like it will actually try to output that coconut image to look like an apple, orange, or banana. Then what you can do is calculate the difference in the pixels between the autoencoder's output image and the input image. If that difference is very high, then you can reasonably assume that the input image was a fruit the model hadn't trained on before. This may sound complex to code, but it's actually very simple. Tensorflow even has a function to get the difference between the two images. This is a tutorial on Autoencoders by TensorFlow: www.tensorflow.org/tutorials/generative/autoencoder You'll see that they are very similar to the CNN that we coded in my tutorial. I hope this helps!
Пікірлер: 42