You can refer to the previous post to learn how to build the object detection model using the TensorFlow Model Maker API and the efficientdet-lite model (https://www.aivision.app/ai/tensorflow-model-maker-part-two/)

Live Demo for Android and EDGE Device

Here is the live demo for Object Detection App on Android and Edge Devices.

Clone the Repo


git clone https://github.com/Abdullamhd/od_efficientdet

Building the Object Detection App for Android

Prerequisites

  • Android Studio
  • Android Device with Android 8 or higher.

Once you have cloned the repo, you can open the project in Android Studio. The project is located in the android_od folder.

Running the App

once you open the project, Wait for the gradle to build the project and then you can run the app on your Android device, make sure the usb debugging is enabled on your device and you have connected it to your computer.

Building the App with your own model

if you built your own model and want to use it in the app, you can replace the model file located in android_od/app/src/main/assets with your own model file. then replace the model name in the file ObjectDetectorHelper.kt in the method setupObjectDetector() with your model name , line 85 , if you built only on model just assign the model name to the variable modelName like this:


        val modelName = "YourModel.tflite"

if you built multiple models, you can use the when statement to assign the model name to the variable modelName like this , assuming you built 5 models


        val modelName =
            when (currentModel) {
                MODEL_EFFICIENTDETV0 -> "Apple_Orange_arch0.tflite"
                MODEL_EFFICIENTDETV1 -> "Apple_Orange_arch1.tflite"
                MODEL_EFFICIENTDETV2 -> "Apple_Orange_arch2.tflite"
                MODEL_EFFICIENTDETV3 -> "Apple_Orange_arch3.tflite"
                else -> "Apple_Orange_arch4.tflite"
            }

make sure to replace the model names with your own model names. then you can build and run the app on your Android device.

Building the Object Detection App for Edge Devices

Prerequisites

  • PC or Raspberry Pi or EDGE Device with Linux OS
  • OS : Windows WSL2 or Linux OS(Debian, Ubuntu, etc..)
  • Python 3.7 or higher

Running the App

assuming you have IP camera or webcam connected to your computer, if you don’t have one, you can use the IP camera simulator app on your phone named “IP Webcam” , search for it on the play store and install it, then run the app , from the app go to menu and select “Start Server”, the app will display the IP address and port number,you can use it to connect to the camera from your computer. then open file detect.py and replace the IP address and port number with your own , on line 22 , 23

IP = '192.168.0.102' # TODO change this to your camera's IP address
PORT = '8080' # TODO change this to your camera's port number
URL = 'h264_ulaw.sdp'
ip_camera_url = 'rtsp://' + IP + ':' + PORT + '/' + URL

then run the following command to install the required packages , make sure you are in the edge_od folder

    pip install -r requirements.txt

then run the following command to run the app

    python detect.py

The dialog box will open with live video feed from the camera , then you can point to the objects you want to detect , in our case we are detecting apples and oranges , the app will display the detected objects with the confidence score with bounding boxes around them , you can also press the q key to quit the app.

Building the App with your own model

if you built your own model and want to use it in the app, you can replace the model file located in edge_od with your own model file. then replace the model name in the file detect.py line 25 with your model name , because i built 5 models, i created a dictionary to store the model names and their corresponding index, in your case you can just assign the model name to the variable selected_model like this:

`

MODEL_PATHS = {
  0 : 'Apple_Orange_arch0.tflite',
  1 : 'Apple_Orange_arch1.tflite',
  2 : 'Apple_Orange_arch2.tflite',
  3 : 'Apple_Orange_arch3.tflite',
  4 : 'Apple_Orange_arch4.tflite',
}

selected_model = MODEL_PATHS[0] # TODO change MODEL_PATHS[0] to your model name

then you can run the app with your own model.

Conclusion

I hope you enjoyed this post, if you have any questions or suggestions, please reach out to me on twitter @Alhadrami