Image Classification App | Teachable Machine + TensorFlow Lite

[@gloriamasengo]

hi, I have an error, when i select images from gallery the app shuts down, what can i do to fix?

[@kianrenzlukeromitares6535]

Is there a possible way that the photo can provide more information such as it's description and benefits and other information?

[@HarshitChaudhary-hc7jo]

Did you find a way to do this?

[@techinfo89]

Please add the option of upload image also. Like when we open the app we should have two options Take photo or upload image from gallery.

[@IJApps]

That sounds like a good idea. Currently, I have a tutorial on selecting images from the gallery that may be helpful: kzbin.info/www/bejne/fmLNkmudq7mneKc Once you get the Bitmap from the gallery, the steps are the same as for using the Bitmap from the camera.

[@shivamkak7981]

Was stuck for 5 hours...helped lots, thx

[@prashantbhatade6367]

Hi can you help mi to creat such kind of app..

[@FRANKANTHUANCERONMALDONADO]

Excellent video, very well explained. I have a problem, I followed all the steps but when I run the application and take a photo, the application closes and does not show me the results, any solution? thank you

[@LuqmanJuzaili]

Hi! Thanks for the great video! May I ask how do I link my android emulator camera to a real life camera such as yours? Mine just show a virtual environment when I tried to test out the scanning part.

[@alexandramariepascual3550]

I'm integrating my LSTM model to the app I'm creating. I followed your code in getting the output. I am getting NaN outputs. What could be the problem?

[@IJApps]

I haven't worked with LSTM TFlite models, so I'm not fully sure. One thing I can think of is making sure that your inputs are correctly processed before going into the model, just like you would have with your model in Python.

[@ainal2888]

hello mr Can the % value of the classification result be displayed in another textview and only show 2 conditions (example: if the value is >90% then the result is object 1 and if the value is

[@ahmadyusupsaepudin672]

yes i want like this too, did you have a solution?

[@johncedricpamigao7811]

did you have solution for this ?

[@hazelgracehilaga314]

What version did you used in android studio sir ?

[@ravereviews6853]

Hey is there anyway to make the recognition live without taking a picture? Like instead of pictures it uses webcam and when it notices the one class plays an audio clip to signify the detection. Let's say you had a big box of mixed fruit, could you get it to notify you when you when the camera comes in contact with an orange in real time? *"ORANGE DETECTED"*

[@IJApps]

I'm not fully sure how to do that, but you can try searching for "android get camera input real time" or "android custom cameraview". This link may help you get started: stackoverflow.com/questions/43091991/android-real-time-get-image-data-from-camera

[@ravereviews6853]

@@IJApps thanks for your help!

[@simranbuvra9832]

which version of android studio you are running? plzz reply

[@ajbalidio.01]

What compiler did you use sir ?

[@VPR_002]

I have one doubt from this video.. How you are try to finding maxpos and confidence..??? How you decide whether this class positions or confidence is give results as apple or banana???? Kindly reply to my answer please😢

[@stanleedoyo2921]

hello sir thanks for the tutorial.. i hope have some code for toast error if you taking a photo not in the model

[@stanleedoyo2921]

but thank you for your code that a big help for my reference source

[@uDubRiceBoy]

Hi, can teachable machine tflite model perform on video real-time detection? Thanks for your help!

[@dentitanium]

I have a question. I've tested my actual model and the TFlite model in Python and they show the same results, but when I deployed it through Android Studio, it shows a different prediction result, it seems I have a problem with the input while playing along with the input code. How can I improve it?

[@IJApps]

Are you showing the exact same image both times? This part on preparing the input is important: kzbin.info/www/bejne/oJmqnmeBecaZgbc Make sure to divide by 255. The full code is available here: github.com/IJ-Apps/Image-Classification-App-with-Teachable-Machine

[@dentitanium]

@@IJApps Yes, I'm using the same images. I'll look into the video further and code you gave me, thank you so much for your response!

[@dentitanium]

@@IJApps Update, we found out that during our model training, we used the Keras preprocess_input function where it converts the RGB sequence into BGR and zero-centers the image. I modified the input loop by changing the red and blue sequence (putFloat blue first, then green, then red) and subtracted the values with its channel mean. It worked! Still, thank you for this tutorial. This helped me for our thesis

[@EYEA6621]

​@@dentitanium Thank you very much, i was follow IJ Apps tutorial and i have same issue with the incorrect result. So i changed the putFloat Blue > Green > Red. Then Successsssssss!

[@srujanhams2443]

Important tensor flow file is not opening to create

[@danilodutra5837]

Excellent initiative! How could I import a .tflite file already created and generated by yolo?

[@eslammansoor4482]

what is deferens between unquant model and quant model

[@amankinde6610]

come up with other videos sir good jop

[@rainbalen7574]

hello sir ay i ask how to apply condition if it is not registered on model it will automatically print "Invalid data"?please help

[@alostsoul9594]

How to use front cam ? Please reply sir....

[@nelsontumbado2518]

i make my fruit identifier but why it only say apricot in classified as

[@서동헌-o5l]

I have question If I take a picture, I want it saved to my develop to my app ui can i this?

[@zaisen2418]

Which algorithm is used for this detection?

[@IJApps]

This is Teachable Machine's FAQ which may answer your question: teachablemachine.withgoogle.com/faq I have a tutorial where we code our own Convolutional Neural Network for image classification on Android: kzbin.info/www/bejne/mJKXY6iPf8lrpMU

[@leolima8585]

How would it be understood to have 3 different types of caches in the same app, user type clear 1 while preserving the second and third, and respectively

[@phunthaoo8247]

can i classify fingerprint? I afraid it is too blur or small so it can't be recognises

[@sajaata2006]

How can I add sound to the result

[@johncedricpamigao7811]

only 3 classes are allowed and can show an accurate output on this project?

[@johncedricpamigao7811]

i hope you can answer my question badly needed :(

[@athukurianilkumar5352]

when i show un trained object it detecting and showing Randoms name which i trained objects name.

[@isabelastateuniversitysanm7268]

greate tutorial! this can help my project for my capstone hehhee

[@jerwingubat6716]

Hello how can i get the real percentage of the detection from the model?

[@arjaybuen998]

have you tried cancelling camera instead of capturing? I just want to ask it because the app that I developed if I cancel or press back instead of capturing, the app immediately close and I need to open the app again. What I want is if cancelled or pressed the back button it will go back to the activity and not closing.

[@IJApps]

Good question. I covered how to account for that in my tutorial on Getting Camera Pictures: kzbin.info/www/bejne/bYLapqVqrbCbY8U

[@armaan_bruh]

Hi , how can we use this in glasses for blind people , so that it can call out the object in front of them?

[@francisluna6577]

hello sir, i have a concern when i was trying to run the and test the code there was an error " cannot find symbol class NonNull"

[@IJApps]

Hello, this StackOverflow answer may be helpful: stackoverflow.com/questions/50470474/cannot-resolve-symbol-nonnull-after-android-studio-update

[@ajbalidio.01]

Where to download tensor flow

[@ivaylomanolov1455]

Hello, how can the teachable machine be trained using text or numbers?

[@kshitijgupta592]

Can I create iOS apps with it?

[@juanandresveravera7952]

can be use for a project in flutter?

[@AICSAnushaBhat]

is this code works if I saved model in quantized format?

[@alostsoul9594]

Bro thanks for your tutorial.. But my image seems to lose its quality can help me fix it?

[@cyrilafedo65]

what is the string name

[@AbdulHannan-xl3vx]

Sir, in your app model is recognized only one image at a time is it possible to recognized two or more images at one time??

[@IJApps]

When it says TensorBiffer.createFixedSize(), change the 1 in the array to how many ever images you want. when you're allocating Direct, multiply by the number of images.

[@healersage2281]

Nice video bro, can you make a real-time object detection the one with bounding boxes and labels in Android studio using TFLite?

[@IJApps]

Sure! It might take some time as I have some other tutorials planned already, but I'll look into it.

[@healersage2281]

@@IJApps Thanks for the answer bro! I have another question, can 2 tflite models possible in one app?

[@moomoo_melody297]

@@IJApps you don't need to add API on it?

[@fxxksxxt4530]

Has anyone made it, I'm having some problems

[@Tan-st5sp]

​@@IJAppshi ... Can I use this code for facial emotion detection???? Plz plz plz rply

[@sabihajahaneasha2939]

I followed every step.But my app crashes after clicking image

[@chamikasesathbandara7338]

Great tutorial. thanks. can we modify it into real time assessing the confidence value?

[@heavengoldiequirimit9940]

hi! i'm facing a problem, when a take the picture and press ok, the app crashes. The error is in the following line: java.lang.RuntimeException: Failure delivering result ResultInfo{who=null, request=1, result=-1, data=Intent { act=inline-data (has extras) }} to activity {com.example.scantest/com.example.scantest.MainActivity}: java.lang.IllegalArgumentException: The size of byte buffer and the shape do not match. Can you tell me why and how do i fix it? tyy :))

[@IJApps]

Hello. What could be happening is the size of your byte array is not large enough to fit your image (maybe you forgot to crop the image). I put the whole project on GitHub: github.com/IJ-Apps/Image-Classification-App-with-Teachable-Machine/blob/main/app/src/main/java/app/ij/mlwithtensorflowlite/MainActivity.java The link above may be helpful.

[@heavengoldiequirimit9940]

@@IJApps thank u soo much! one more question, how do i add some code that the picture is from gallery? bcuz i tried, but it wont work and i dont know how.

[@IJApps]

@@heavengoldiequirimit9940 I have a tutorial on selecting images from the gallery here: kzbin.info/www/bejne/fmLNkmudq7mneKc

[@danielgomez1194]

Hi! I m facing the exact same problem, I m using another model to classify other stuff , I erased the fruits model and uploaded my model with the same name, has the same sizes (224,224,3) and y changed the labels to 6 different classes , oh and its also float32 for efficientnet... Did you fix that error?? I ve been trying for days and dont know what else to do

[@danielgomez1194]

Never mind, I find out my app needed some extra dependencies in the build.grandle for my model

[@user-cb3ge1zr5t]

Hello, I am having trouble in the confidences array cause it has only 1 value but I have 2 classes to classify.

[@IJApps]

In the case of binary classification, it may be that if your output is closer to 0.0, it is one class. If it is closer to 1.0, then it is the 2nd class. Try seeing if this works.

[@localbus7008]

Sir when I capture the picture, after that app closed it say app experiencely crashing ..? Why I am facing this issue?

[@Haleymrn]

I have the same issue, have you figured it out now?

[@localbus7008]

@@Haleymrn not yet

[@localbus7008]

@@Haleymrn waiting for sir's reply

[@IJApps]

@@localbus7008 I need more details about the problem. When you go to Logcat in Android Studio, it typically shows you where the program crashed (what line number in your code). That may be useful to you.

[@localbus7008]

@@IJApps OK sir I will check and then I will tell you here...

[@mbuguanjane5612]

I have always gotten scared when comes to Artificial Intelligence but this seem straight forward

[@IJApps]

Glad to hear you found this tutorial approachable and easy to follow! I have more AI videos here: kzbin.info/aero/PLLmkb5CTw5rSEpCgictoZgfSYZ_F0o0IQ

[@kevovoke]

How do you handle when an object doesn't have a class? Like an object you didn't train it with like a car for example.

[@IJApps]

Two ways I can think of off the top of my head are to have another class containing an assortment of images that don't belong to any of the other classes. You could also try doing something where you check the confidences array returned by the model and only display a category if the maximum confidence is above a certain threshold

[@IJApps]

I'm sure there'd be other recommended methods from a Google Search.

[@fatihpc]

Hello there. In which folder do I put my "model.tflite" file and it works fine?

[@IJApps]

Hello. Sorry for the late response. You can check the video description for my GitHub repo.

[@joenamarkmana-ay3797]

My program is running, but when I take a picture and when I ok the image it turn automatically exit. May I ask if what happened to this kind of error? Thank you

[@IJApps]

Android Studio has a very useful Logcat feature in the bottom left. Go to Logcat while your device is connected to the laptop, and you'll see the error when the app crashes. It will also show the line #.

[@shreyass9915]

sir i facing this error: import androidx.annotation.NonNull; ^ symbol: class NonNull location: package androidx.annotation

[@IJApps]

androidx is a new version of the Android support library. You can try just making it android or search up Android documentation nonnull and see how they imported it.

[@envy_ez4me]

how can I modify the project such that it will not take any picture. Just the object will be kept in front of camera and it will predict the same. Please help.

[@IJApps]

You can try searching for this: android studio get camera realtime Maybe this can help: hamzaasif-mobileml.medium.com/getting-frames-of-live-camera-footage-as-bitmaps-in-android-using-camera2-api-kotlin-40ba8d3afc76

[@antonizheng4521]

the confidence is out of bound, why? i have 60 dataset in model

[@bollynerds8358]

Bro your main activity.java is incomplete plzz give complete file

[@Carolinesherina]

can you do it kotlin version?

[@athukurianilkumar5352]

in your video your showing trained objects only if you show car or bike etc.. object what is output of it.

[@IJApps]

The model has learned to output 3 different classes: apple, banana, and orange. As such, when it receives an image, whether or not it is one of those 3, it will output one of those 3 labels.

[@johncedricpamigao7811]

@@IJApps you know how to prevent it from happening ?

[@josuavalentino4159]

Nice Work!!! but why when I take the pic the app will forced close.. I'm using my Redmi Note 8 for the emulator

[@IJApps]

If you go to Logcat along the bottom of Android Studio and rerun the app, you should be able to see the error when it crashes and what line of code causes it. If you still can't fix the error after viewing Logcat, you can copy and paste the error message here.

[@Haleymrn]

Hi, what if I have two models uploaded in the folder. How do I specifically modify the code so it can detect which model I am using?

[@IJApps]

Based on your mode's names, you would use different things. Ex: your model is model.tflite. In the code, that's why we use Model.newInstance and Model.Outputs It all depends on what you named your two models.

[@Haleymrn]

@@IJApps found it. thanks! Btw, I have questions. What happens if we do not add the additional codes in the input like bytebuffer? Will it still work? Do you have available code for the inputs for object detection? Thanks!

[@janphilipdelacruz621]

sir the apps is running but when i take a picture and return to app it will shut down i dont know why

[@IJApps]

You can try looking at Logcat in the bottom left of Android Studio. It will tell you what error is occurring and where. Make sure to keep your device connected to Android Studio and it should worl

[@mohsinjaved9574]

I use teachable machine but it can't working in my browser not showing camera when clicking on button

[@IJApps]

You may not have given the website permission to access your laptop's camera. You can check your browser's permissions for that site. If you're on Chrome, at the right side of the search bar there will probably be an icon for the camera.

[@YogaPratama-xz3dz]

I tried to export the Tensorflow Lite model with the Quantized option but after waiting for the process the message "something went wrong while converting" and the downloaded rar file could not be opened, why is that?

[@IJApps]

The error seems to be with Google's Teachable Machine. The Teachable Machine page has to be left open while the model is being converted. I'm not sure what else could have caused the issue.

[@ainal2888]

I have followed the video step by step, but when the application is run, the label / is classified as not the same as the confidence value, the highest value should be the label, can you help me, thank you

[@IJApps]

Hello. Have you checked if your String[] classes array has the class labels in the same exact order as when you were training the model on the Teachable Machine website? That might be why that's happening. Here's my code on GitHub that you can check out: github.com/IJ-Apps/Image-Classification-App-with-Teachable-Machine

[@josephmercado7613]

I am new to programming, I made 20 classes on teachable machine, I would like to show only 5 confidences with the highest confidence, can you help me? Thank you!

[@stephansozkes5082]

Nice video bro, what I've been searching for. Only problem is it's successfully working until I take the photo when I take the photo and click ok the app stops running. Could you please help?

[@IJApps]

What errors are you getting when the app stops? If you open Logcat in Android Studio it should show you the errors when the app crashes

[@janjobertcortez9478]

does this app need internet connection to worK?

[@IJApps]

No! The model runs locally.

[@kuyalecpogi]

can you make another video of it working in real time?

[@swayamparte4417]

love you bro! thanks a lot your video is the reason my project is alive!

[@sahilpatil2992]

can you help me out? after clicking the image and hitting my app just crashes

[@mikemanahon2696]

how to modify code and choose image from gallery?

[@IJApps]

I have a tutorial on choosing images from gallery. The rest of the code should be the same: kzbin.info/www/bejne/fmLNkmudq7mneKc

[@juliuszamora9082]

Good day, I having a problem when taking a picture App will stopped working. I have also Warning : Call requires API level 23 (current min is 21): `android.content.ContextWrapper#checkSelfPermission`

[@muhammadarsalan6072]

please help me sir i am facing problem while downloading quantize.tflite file. It says something went wrong on the could. And the downloaded file is corruped and unknown format..

[@IJApps]

Make sure to not leave the Teachable Machine tab while the model is training. Also make sure you have a stable internet connection.

[@muhammadarsalan6072]

@@IJApps thanks for replying brother i have stable internet connection and also i donn't leave the tab while training but its gives me error that is "Something go wrong on the cloud". I stuck on that since from last 6 days. Teachable machine works well when i made 10 to 20 classes but when i make more than 20 classes for example 30 classes its not working.

[@IJApps]

@@muhammadarsalan6072 Maybe they have a limit on the # of classes because it is too computationally intensive. Or maybe an error just occurs. I've never tested TeachableMachine for more than 5 or 6 classes.

[@ahmadyusupsaepudin672]

hello mr I want to make if the confidience

[@IJApps]

When you get the maximum confidence in the array, have an if statement to check if it is less than 50%. If it's less than 50, then display "no identification"

[@mmaalielhaj5043]

I am a beginner in learning Python and mobile applications please, Can I get this project as an open file to learn from?

[@mmaalielhaj5043]

I couldn't run the code to work please, I want it as files and I will open it to learn

[@IJApps]

​@@mmaalielhaj5043 The code is available on GitHub: github.com/IJ-Apps/Image-Classification-App-with-Teachable-Machine Typically, I put the link for the tutorial code in the video description. I hope this helps.

[@mmaalielhaj5043]

@@IJApps thank you so much 👍

[@autopilot_youtube]

Thank you! Отличный урок!

[@hasim5688]

Thank you so much. We want more video like this.

[@KasanggaVlogs]

Awesome tutorial! Do you have a kotlin version of this, thank you so much for your help!

[@anotherrubi9885]

Great video and great explanation! This is of so much help.

[@agytfreefire]

Hello😭😭 please solve my problem when I click to take picture and click on tick it crash

[@IJApps]

Along the bottom of Android Studio there is a tab that say Logcat. Select Logcat. When logcat opens up there will be a dropdown to select which log you want to see. Select "Error" from the dropdown. You'll probably see a lot of red text. Rerun the app. Click on take picture. This time when the app crashes on your laptop, logcat in Android Studio will show you why the error occurred and what line it occurred at. You can use this to fix the problem.

[@agytfreefire]

I have fixed when I remove orange banana and more from string and add names of the given in model then it works

[@parthkumar19]

Great Video. Thanks. Keep it coming.

[@rrqhoshi5140]

request using kotlin sirr

[@ct1egh]

Great tutorial. You got a new subscriber 👍 Thank you !

[@MELO-ij8wy]

Great video, Thanks a lot for the tutorial

[@aitorsanchez2581]

Thanks! Very useful content

[@janhenripang-es9728]

How to display the confidence of only 1 class?

[@IJApps]

Use the setText function and the confidences array. "setText(confidences[0])"

[@janhenripang-es9728]

@@IJApps thanks! but how can I turn into percentage?

[@IJApps]

@@janhenripang-es9728 Multiply by 100. It's all in the video. 13:00

[@masleniamubarrat2606]

hei, needed some help.

[@agytfreefire]

I beg you how to remove confidence

[@IJApps]

What do you mean by this? If you don't want to display the confidences simply remove the TextViews.

[@agytfreefire]

When I remove text views it shows error

[@IJApps]

@@agytfreefire You have to remove all code associated with the textViews. So, both the TextViews in the XML file, and then in the Java file you have to remove the findviewByID() for the textview, and all the places you use the TextView variable.

[@carvajalflorenciacarlosseb7503]

Thank you very much, it was a great help 😊😊

[@mohammadridhopardomuan2486]

amazing tutorial, big thanks my man

[@hominhquyen1192]

in my project confidences.length = 197754 . in line if(confidences[i] > maxConfidences){ maxConfidences = confidences[i]; maxPos = i; } i = 195834 In line result.setText(classes[maxPos]); i see classes.length = 3 ( classes = { "other", "vietinbank" , "vietcombank" } ==> out of range. pls help me

[@IJApps]

You may not have coded the input part correctly. You can check out the full code at this link: github.com/IJ-Apps/Image-Classification-App-with-Teachable-Machine

[@hominhquyen1192]

@@IJApps when i change model.tflite with metadata. It work for me.

[@henzzhenzz9015]

Sir, how to get the x and y bounded box dimension of the detected object in the image classification? is it in the outputFeature0? just like object detection bounded boxes but for image classification. I want to get the height and the width of the box detected in the object detection using only the image classification. I need it for the sorting of sizes of the object neither it is small, medium, or large and i want to sort it using the bounded box detection dimension sizes. Thank you if you reply this comment.. :)

[@IJApps]

Because this is an image classification model, we can only get the class of the image. for getting the coordinates, you need an object detection model.

[@alpapskie25]

This Guy is amazing!

[@JoloMovies-lr2rz]

too good explain method , woop

[@muhammedtalhayilmaz3069]

Thank you sir

[@jindhu2608]

Superb

[@imadsaddik]

Thank you so much

[@akmalyafi1470]

Thank You!

[@m.abdullah482]

This is amazing

[@HerrHafiz]

nice i love it