Deploy yolo model android | studio | yolo v5 | nms | java | tflite

Hi in this video you will learn how to deploy yolo v5 model in your android app using tflite, This is very step by step video explaining, exactly how to integrate android app using java language with the yolo v5 object detection model. We start by building very simple ui, then proceed to write up backend code for making prediction using our yolov5 tflite model, we also see how to convert the model from pytorch format into tensorflow lite format easily and quickly.

In this comprehensive tutorial, learn the step-by-step process of deploying a YOLO model on an Android device using Java. I cover everything from building the UI to converting a PyTorch model to a TensorFlow Lite model. Follow along as I guide you through writing labels, loading the model, implementing the Non-Maximum Suppression (NMS) method, and drawing predictions on the canvas. By the end, you'll have a fully functional YOLO model on your Android device, ready to showcase to users. Don't miss out on this hands-on guide for practical implementation.

—————————

—————————

Connect with me:

yolo,yolov5,deploy,android,java,deploy yolo model,tflite,tensorflow,step by step,guide,android studio,deploy ml model,deploy object detection android,android ml,android deep learning,tensorflow lite

[wolpumba]

*Summary*
*Introduction and Project Setup*
- 0:00 - Introduction to deploying YOLO V5 model on Android using Java.
- 0:11 - Steps to create a new Android project using the latest version of Android Studio (Giraff version).
- 0:27 - Project named 'yolo deploy', using Java language and Android 8.
- 0:35 - UI changes in Android Studio Giraff version and enabling the new UI in settings.

*Designing the App Interface*
- 0:52 - Designing the app UI with an ImageView and a prediction button in a relative layout.
- 1:40 - Setting onClick functions for image selection and prediction.

*Implementing Image Selection and Bitmap Handling*
- 2:08 - Defining the 'select image' and 'predict' functions in MainActivity.
- 2:41 - Implementing image selection from device storage using intents.
- 3:07 - Code for image picking and bitmap handling in Android.

*App Testing and Troubleshooting*
- 4:30 - Running the app and troubleshooting a compile SDK version error.
- 5:02 - Testing the image selection functionality in the app.

*Preparing YOLO Model for Android*
- 5:45 - Outline for loading and converting YOLO model to TensorFlow Lite format.
- 6:01 - Cloning the YOLO V5 repository for model conversion.
- 7:04 - Converting the PT model to TensorFlow Lite using a Python script. Visualization with Netron app.
- 8:01 - Overview of the YOLO model output structure and Non-Max Suppression (NMS) algorithm.
- 10:05 - TensorFlow Lite doesn't come with NMS.
- 10:33 - Referring to a third-party GitHub repository for NMS implementation.
- 11:08 - Adding the converted YOLO model to the Android project assets folder.

*Setting Up Model Assets and Java Classes*
- 11:26 - Adding a `labels.txt` file in the assets folder, listing classes the model is trained on.
- 12:09 - In Android Studio, creating a new Java class named `yolo V5 tflight detector`.
- 12:36 - Creating another Java class called `recognition`.
- 12:50 - Adding TensorFlow Lite dependencies in the module-level build.gradle file.

*Configuring and Utilizing the YOLO Model*
- 13:25 - Making changes in YOLO V5 class for image and output sizes according to the model.
- 14:37 - Defining and instantiating the YOLO TensorFlow Lite detector in `MainActivity`.
- 15:08 - Initializing the YOLO model and setting up the model file path.
- 15:16 - Implementing the detect function on the predict button click, processing the bitmap.
- 15:49 - Utilizing the recognition object from the recognition class to get detection details.
- 16:21 - Converting bitmap to a mutable bitmap and setting up a canvas.

*Finalizing and Testing the App*
- 17:00 - Drawing detection rectangles and labels on the canvas if confidence is above a threshold.
- 18:07 - Displaying the processed bitmap with detections in the ImageView.
- 18:19 - Running the app and testing the image loading and prediction functionality.

Disclaimer: I utilized ChatGPT 4 (11 Nov 23) to condense the video
transcript into a summary. I employed Prompt 1 to generate
timestamped bullet lists. I then used Prompt 2 to organize these lists
into sections with titles. The summary was manually formatted using
YouTube comment markup.

Prompt 1: "Generate a bullet list summary, including starting
timestamps for each point."

Prompt 2: "Split the following bullet list into sections. Create
section titles. Keep timestamps for the bullets. Use this format for
titles: *title*."

[sanjoetv]

can you make a video about yolov8 please
how to integrate it in android studio

[lykagracetabiolo]

This is a great help! Been struggling finding videos of YOLO deployment.👏

Can you do also real-time/ live detection using phone camera?

[terrylimyl]

Great Video, hopefully you can do onefor real-time object detection.

[varunr]

Hai I am getting alot of bounding box with improper coordinates and also detection results are not good.what change can i do ?

[tomlai]

Hi, how come it wont work if i use another tflite file like yolov8? i changed the 6300 and the imagesize accordingly as well

[aihtishamarfi]

Sir make a video on if human is detected perform some operation show Toast etc
Please sir i need

[ahmadsarji]

Sir please can you make a video on real time object detection using yolo
I need it for my graduation project and the deadline after 6 days only

[mahrukhhafeez]

In Android Studio the targetSdk cannot be equal be 34. It gives error. it doesn't give error on 33. Should i keep it the same?

[Amlosiorh]

Should i change label. Txt if i using my custom yolov5?

[chandrashekarreddy]

How to create '.pt' file and also in your repository which file need to be converted into .pt file. Please explain

[marshalmochi]

Hi, thanks this is very formative.

Can you create a YoloV8, torch.load(Model.pt) to detect and predict a cat in a video? for a web flask python deployment

[unisimplex]

hey Fantastic video, i also want to count number of objects of a class detected and show it to user, how can i do it ?

[Mjnkk]

Can you do object detection app that include capture and load image using yolo8. tks u so much

[lullmzc]

Do you also know how to deploy trained yolov7 on Android using Java and how to do the detection?

[nr_developer]

Sir did you make a app to capture 360degree image and show it in another activity . Bro it is one activity of my college project i have just 1 months i mean i need to submit my project in January so if you help me regarding that then i am so helpful to you please make a video there are no video on the internet to capture 360 degree image and preview it in another activity .... There are no video regarding that ... IF you make any video regarding that please informe me

[bingebox]

This Algorithm for NMS, can it be used for a yolov8 model?

[uzqepft]

Hey bro im new to android development can u please tell me where to start and wht to learn. Please help

[mohd.abdulghani]

hai, can you make a video tutorial how to deploy yolov5 instance segmentation on android? thank you

[sanjoetv]

i try to use a yolov5 model which has an input size of 640x640 and an output size of 1, 25200, 6

but their is no prediction being displayed.

even i change this values

private final Size INPNUT_SIZE = new Size(640, 640);
private final int[] OUTPUT_SIZE = new int[]{1, 25200, 6};


to match what i needed can you please help me why their is no prediction being displayed