@@CodeWithAarohi while we were working on our project we trained and quantized our model. it runs at 400ms but we need it to be faster. we trained our model with yolov8 and converted this model to tflite. how can we make it faster, what would you recommend?
hello maam, thank you so much for this. I implemented this using my custom dataset but i cant seem to figure out how to erase the bounding boxes when no objects are detected. can you help me? thank you!
Awesome, very helpful!, I have a question: If I would like to recognize a Pothole road with my App and also capture the GPS location, and store it in a table how can I do That in an android app? I am thinking in a car with a gopro camera connect to the android, while im driving i am detecting the quality of the street.
when I tried to export to tflite, i got error ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed.
from ultralytics import YOLO # Load a model model = YOLO('yolov8n.pt') # load an official model model = YOLO('path/to/best.pt') # load a custom trained model # Export the model model.export(format='tflite')
Android Studio recommended it. When creating an app in Android Studio, there's an option to choose between Java and Kotlin. Since "recommended" is written next to Kotlin, I chose it. Additionally, some of my subscribers requested that I work with Kotlin :)
hello, any suggestion how to optimise the app further so that the ms and the detection becomes faster? and maybe how to include int8 format? any links/suggestions/example projects on how to make the detection faster? thanks for any answer in advance
To create a standalone app icon on your Android phone , you need to package your app into an APK file and install it on your phone. This way, you won't need to connect your phone to your PC every time you want to test the app.
I dont have. But the process will be same for custom dataset also. Just train your object detection model then convert it to tflite and use that tflite model in your android app.
@@CodeWithAarohi Thanks, What I mean is that Do you have a ready-made models or datasets with various images? So that, I have variety of images than i can scan? Not only the fork and such?
hello maam , i cant convert the the dataset to tflite file , it cause my tensorflow version is not 1.13.1 , but when i want to pip install it , it says cannot find the tensorflow version , how can i fix this
Dear Madam Aarohi, Thank you very much for your contribution. I would like to know how I could complement the external annotations to the yolo model, for example 'supervision' annotations that set an object counter? The process you show also run in yolov9 right? I mean app processing
hi ma'am, may I ask something, when I try to pip install tensorflow 2.13.1 inside jupyter. This error message appear: Could not find a version that satisfies the requirement tensorflow==2.13.1 (from versions: 2.16.0rc0, 2.16.1), Do you know what caused it and how to fix it? Thanks
This means that the pip version you are using don't have tensorflow 2.13.1. Now, to solve this issue there are 2 methods. 1st method is to downgrade the pip version and install tensorflow 2.13.1. 2nd method- Just install tensorflow 2.16.1 with your current pip version.
Thank you very much for this video 💕Could the next video be about creating an audio feedback for this model using Spread-to-Text libraries etc. by Android Studio?🤭🙏
thanks for the tutorial, it was very helpful, I have a question, the model I created is running quite slow, it was much faster when I tried it directly on the yolov8 model in python, is this slowdown because I converted it to tflite or do I need to make improvements in the code
You can quantize your model. quantization can reduce the model size and improve inference speed. Also, make sure that the input data format matches what the model expects.
yolov8_env folder is a folder which I have created myself. You create that folder yourself and then work under that folder. And don't say sorry and don't worry if a question is stupid or not, Important thing is you are trying and learning. Good Luck!🙂
Hello mam. I have a YOLOv8 model for detecting 10 object. When I change the path with my model and labels the app crushes. your app is build to detect 4 object. what should i do to deploy a 10 object detecting model in your code?
Hello Aarohi, How can we display its percentage/size similar to how it's shown in a GitHub image, like 'spoon 0.92' or 'fork 0.84'?" Currently, as shown in the video, it only displays 'spoon' and 'fork' without the percentages.
How to make the bounding box gone after a while? Can this use for yolov5 model? How to make detection more faster? How to show confident score beside class name?
override fun onEmptyDetect() { binding.overlay.clear() } and modify clear (which is dead code anyway) to fun clear() { results = emptyList(); invalidate() } and add this if to draw() after the call to the super class if (results.isEmpty()) { canvas.drawColor(Color.TRANSPARENT) return}
Mam you have made plenty of videos on yolo model Mam if possible could make a video human action recognition using lstm cnn and mediapipe Also make it end to end that would be great
@@CodeWithAarohi thanks mam that issue is resolved but now i'm unable to find the data.yaml file from the github repo neither i am able to find the dataset which you used... is it not publicly available?
Hi, first of all thank u for this amazing content. I want to ask you a question. I am trying to create a model based on YoloV8 to detect the color of the traffic light. I am having an issue that when I use the train model with the camera in video mode, the model takes a while but detect the traffic light but when I try to detect only by sending one frame, the detection is null. I do not understand why when you send multiple frames it gives you a better detection. Could you help me with this? Thank u
Hello! I'm here again. Not related. But is it possible to build an application or to train models with gesture recognition? I already made a static sign language translator but i want it to be dyamic. It can identify dynamic sign language. Is it possible? If yes, then how?
hi great job thank you for the video .... plz i need to be part of your memeber channel i need to learn more from you .. and have full folder of code plz help me
I got an error "ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. " when I tried to export to tflite
You are such an amazing content provider for Professionals working in Artificial Intelligence & Data Science
Thank you so much
My pleasure!
For long time waiting for this, thanks
You're welcome! I'm glad I could help 🙂
@@CodeWithAarohi Thank you very much Ms. Aarohi 🤍
EVery helpful & simple way of explaination
Thanks!
just show this video and got to know another legendary creator with low sub counts
Glad my video is helpful!
thank you maam for this video tutorial this will help me a lot in my new project
You are welcome 😊
YOU ARE BRILLIANT, BEAUTIFUL, THANK YOU SO MUCH. U SAVED ME.
Glad I could help!
I am curious to know the inference speed of the model on your device. what is the highest FPS you have acheived?
thank you for this video. helped me and my friends so much you wouldnt even believe. i wish you best. i couldnt thank you enough
Glad to hear that! Good Luck 😊
@@CodeWithAarohi while we were working on our project we trained and quantized our model. it runs at 400ms but we need it to be faster. we trained our model with yolov8 and converted this model to tflite. how can we make it faster, what would you recommend?
merhaba dostum acaba bu konuda proje yapıcam acaba senden veya başkasında faydalanabileceğim bir link veya başka github bırakabilirmisin teşekkürler
hello maam, thank you so much for this. I implemented this using my custom dataset but i cant seem to figure out how to erase the bounding boxes when no objects are detected. can you help me? thank you!
You are a true inspiration madam.Thanks a lot for the lessons.
Thank you so much for your kind words! I'm truly honored to have been able to inspire you🙂
Awesome, very helpful!, I have a question: If I would like to recognize a Pothole road with my App and also capture the GPS location, and store it in a table how can I do That in an android app? I am thinking in a car with a gopro camera connect to the android, while im driving i am detecting the quality of the street.
Great video. Can I ask what did you use to annotate your dataset?
LabelImg tool. You can install it by writing - pip install labelimg
Will you create video on improving the model?
Noted!
when I tried to export to tflite, i got error
ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed.
Great Work, would be very helpful
Thank you!
Hello, I currently have a trained model in roboflow. How to integrate the model to an application by flutter?
Very informative thanks for sharing
You're welcome! I'm glad you found the information informative.
Impressive Content
Thanks!
Thank you .. i want ask u .. How long did it take? and Is this command correct for conversion in cmd (yolo export model=yolov8n.pt format=tflite) ..
from ultralytics import YOLO
# Load a model
model = YOLO('yolov8n.pt') # load an official model
model = YOLO('path/to/best.pt') # load a custom trained model
# Export the model
model.export(format='tflite')
Where can I get the sample training dataset and the data.yaml file? Alternately, I would need the format of the label files.
docs.ultralytics.com/datasets/detect/#ultralytics-yolo-format
You are very good at what you do… trust me! After deploying it on android. How can I share the application or how someone else could get it
Generate APK file and then upload the app on google play store.
thanks for the video. Can I integrated this app to real time camera feed from a wireless camera
Yes you can!
Thanks a lot for the video. I have a question. @CodeWithAarohi why did you use kotlin instead of java like in yolov5 android app?
Android Studio recommended it. When creating an app in Android Studio, there's an option to choose between Java and Kotlin. Since "recommended" is written next to Kotlin, I chose it. Additionally, some of my subscribers requested that I work with Kotlin :)
hello, any suggestion how to optimise the app further so that the ms and the detection becomes faster? and maybe how to include int8 format?
any links/suggestions/example projects on how to make the detection faster?
thanks for any answer in advance
How can I lower the MS. Currently on my phone it is around 800MS wich makes the detections have a significant delay
Use a better mobile or quntaize the model
thanks for sharing. Every time we have to connect the mobile phone with PC or is there any separate app icon is created on phone to run ?
To create a standalone app icon on your Android phone , you need to package your app into an APK file and install it on your phone. This way, you won't need to connect your phone to your PC every time you want to test the app.
You're such an amazing professional. My new IDOL for now! Anyways, Did u have a not custom dataset. Its like its already made for tflite?
I dont have. But the process will be same for custom dataset also. Just train your object detection model then convert it to tflite and use that tflite model in your android app.
@@CodeWithAarohi Thanks, What I mean is that Do you have a ready-made models or datasets with various images? So that, I have variety of images than i can scan? Not only the fork and such?
@pinocchio200 sorry, I don’t have. But you can use pretrained model like the pretrained model train on coco dataset which have 80 classes.
hello maam , i cant convert the the dataset to tflite file , it cause my tensorflow version is not 1.13.1 , but when i want to pip install it , it says cannot find the tensorflow version , how can i fix this
You need to convert your trained model to tflite not dataset. To install temsorflow, just run pip install tensorflow
Dear Madam Aarohi, Thank you very much for your contribution. I would like to know how I could complement the external annotations to the yolo model, for example 'supervision' annotations that set an object counter? The process you show also run in yolov9 right? I mean app processing
Yes, but you need to make the changes to the code which is related to object detection model in order to fetch the detections.
ok thanks Madam, please when you have a time free you make a video ussing yolo and supervision for android. Thanks.
how can i put the confidence of the image detect on the screen (like for example "label : confidance")?
Thank you and great work
You can make changes in the file which is responsible for putting bounding boxes.
Will you make any video for attention mechanism? (Like what you made for resnet50, that was amazing)
I will try!
hi ma'am, may I ask something, when I try to pip install tensorflow 2.13.1 inside jupyter. This error message appear: Could not find a version that satisfies the requirement tensorflow==2.13.1 (from versions: 2.16.0rc0, 2.16.1),
Do you know what caused it and how to fix it? Thanks
This means that the pip version you are using don't have tensorflow 2.13.1. Now, to solve this issue there are 2 methods. 1st method is to downgrade the pip version and install tensorflow 2.13.1. 2nd method- Just install tensorflow 2.16.1 with your current pip version.
Hi maam,
How did you install cuda and torch ??
Can You please explain or If you have already video please share it here.?
Check this video from timestamp 4:00 ruclips.net/video/A-rzknbjp5M/видео.html
Thank you very much for this video 💕Could the next video be about creating an audio feedback for this model using Spread-to-Text libraries etc. by Android Studio?🤭🙏
Thank You!
You're welcome!
i cant run the code in android studio, please help
Thanks a lot Aarohi
Welcome!
Add Configuration is empty , then can't run. How to fix ?
Miss, where is the confidence score? 🙃 I need that
You can modify the code
Please guide me, I know nothing about android studio🙏🏼
thank you so much
Welcome 😊
Great job as always! Thanks Aarohi. Do you sell online training?
No
thanks for the tutorial, it was very helpful, I have a question, the model I created is running quite slow, it was much faster when I tried it directly on the yolov8 model in python, is this slowdown because I converted it to tflite or do I need to make improvements in the code
You can quantize your model. quantization can reduce the model size and improve inference speed. Also, make sure that the input data format matches what the model expects.
Greeting Ma'am, can the YOLOv9 be used to classify the images as i am trying to build a project on image classification . Thank You
No
thank you for the helpful tutorial. Would you consider doing one for object counting as an android app ?
Sure thing!
Hello ma'am, can we use object detection and object classification at a same time?
Yes, you can. You can combine the results of both the models and display on each frame of video.
where can i find the files in the yolov8_env folder to do the training... sorry for probably stupid question... thx
yolov8_env folder is a folder which I have created myself. You create that folder yourself and then work under that folder. And don't say sorry and don't worry if a question is stupid or not, Important thing is you are trying and learning. Good Luck!🙂
Does the int8 quantized TFLite file work with this app?
I haven’t tested that but I think it will work.
Hello mam. I have a YOLOv8 model for detecting 10 object. When I change the path with my model and labels the app crushes. your app is build to detect 4 object. what should i do to deploy a 10 object detecting model in your code?
Make sure that you have done all the changes as per your custom dataset. In this video, I have explained all the changes you need to make.
Hello Aarohi,
How can we display its percentage/size similar to how it's shown in a GitHub image, like 'spoon 0.92' or 'fork 0.84'?"
Currently, as shown in the video, it only displays 'spoon' and 'fork' without the percentages.
You need to print the confidence scores also.
really very helpful.
Can you please upload image segmentation on android app?
Sure
Can you please make new video using flutter with yolo11
I will try
How to make the bounding box gone after a while? Can this use for yolov5 model? How to make detection more faster? How to show confident score beside class name?
override fun onEmptyDetect() { binding.overlay.clear() }
and modify clear (which is dead code anyway) to fun clear() { results = emptyList(); invalidate() }
and add this if to draw() after the call to the super class
if (results.isEmpty()) {
canvas.drawColor(Color.TRANSPARENT)
return}
Will this work with yolov9
Haven't tried yet! Will test soon and let you know.
Thankyou so much maam
Welcome!
Is Yolo object detection offline or internet required?
offline
i am unable to run the app, "This app has stopped" did anyone know why is this happen? I use android 7 device
The github code does not match the one shown in the video. No: step 2.... line in the code
same here
When I try to train the model in vscode using gpu. Why it became hang my laptop?
If your GPU doesn't have enough memory to handle the model and the training data, it can lead to hanging or crashing.
Can I use Android Studio Koala?
could you please create YOLOv8 android app that use onnx
Hi,
Can we train the model on Mac's with M series?
Yes, you can train although they may not match the speed of dedicated GPUs like those from NVIDIA.
Mam you have made plenty of videos on yolo model
Mam if possible could make a video human action recognition using lstm cnn and mediapipe
Also make it end to end that would be great
I will try!
thank you safe my freelacer project
You're welcome! I'm glad I could help 🙂
what is your python version? I tried with 3.9 and 3.11 and an error occurred when exporting to tflite
Python 3.11.9
@@CodeWithAarohi Thanks, I skipped the metrics part, I don't know if that was the problem, but it worked
for fine tuning the model mam you took 2 GPUs but its throwing me an error when i run that on my machine
i dont have 2 GPUs
write --device 0 for 1 gpu
@@CodeWithAarohi thanks mam that issue is resolved but now i'm unable to find the data.yaml file from the github repo
neither i am able to find the dataset which you used... is it not publicly available?
@@syedmuazuddin8721 you can get the dataset.yaml file from github.com/AarohiSingla/yolov5
And you can download dataset from roboflow universe
Can I ask where the step_2-android_app ? because I count find it Plz answer me
github.com/AarohiSingla/Object-Detection-Android-App/tree/main/android_app/android_app
Hi, first of all thank u for this amazing content. I want to ask you a question. I am trying to create a model based on YoloV8 to detect the color of the traffic light. I am having an issue that when I use the train model with the camera in video mode, the model takes a while but detect the traffic light but when I try to detect only by sending one frame, the detection is null. I do not understand why when you send multiple frames it gives you a better detection. Could you help me with this?
Thank u
for detecting color is easier to get the color of the traffic light.
getting error when trying to run the .pynb file.
Hello! I'm here again. Not related. But is it possible to build an application or to train models with gesture recognition? I already made a static sign language translator but i want it to be dyamic. It can identify dynamic sign language. Is it possible? If yes, then how?
Yes, you can do that. For that, train your model on sign language gestures. Then integrate the model into an application.
@@CodeWithAarohi What tech should I use? Do you have references where i can visit?
do you have a full application code for this with UI?
Yes
Is there a way to lower the MS maam? Can you teach us. Pleaseee
Sure
can you do a react native integration with live feed camera in the two weeks that will helps a lot thank you
I never worked with react native. So, not sure!
Where did you provide the dataset ?
You can provided any computer vision dataset since Roboflow
can you do keypoints detection please? using yolov8 on android
Sure!
Thanks you very much
You are welcome
thankyouuuu🙏🙏🙏🙏🙏🙏
Welcome 😊
please make a video for yolov5 in kotlin
Sure
@@CodeWithAarohi thanks you 🙏
i need its documentation
I don't have any documentation but you can see the documentation on yolov8 here: docs.ultralytics.com/
thanks!
Welcome!
Mam can you start Deepstream
Sure!
app code is there in repository?
yes
hi great job thank you for the video .... plz i need to be part of your memeber channel i need to learn more from you .. and have full folder of code plz help me
Yolov5 to kotlin realtime🙏
I will try
I got an error "ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. " when I tried to export to tflite