Many companies invest in how to create image recognition application with android. To prove that they have a best machine vision.
The Machine Vision In Android apps
Because rapidly growing up image recognition in the market. Many industries creating image recognition.
Therefore, let’s start on how to create image recognition in android using TensorFlow. First of all, let tackle TensorFlow. Here is a quick introduction for you.
Using TensorFlow To Create Image Recognition Application In Android
TensorFlow is a library from google that can provide deep learning methods. Such as the following
- linear models
- regression
- image recognition
- audio recognition
- and etc..
Because of his popular TensorFlow constantly updated. The library is available in java, python, and has two mobile variants. Such as TensorFlow mobile and lite.
These are the evolution of the Tensorflow mobile but does not support all the purposes. Listed below are the requirements for Tensorflow:
- The Dependency of TensorFlow preview
- Initialization of Tensorflow
- TensorFlow image classifier
- Classifier interface
So, let’s start!
Create Image Recognition Application With Android
Therefore I show it to you the basic step on how to create an application with android.
- The first thing to do. Build a new android project. Also, choose an empty activity and name it a “Classifier Activity”
- The second step is in your
–> build.gradle next is –> add compile org.tensorflow:TensorFlow-android:1.2.0-preview’ in dependencies.
- The third step is to create an asset folder in the app folder. Example -> right-click on app -> new -> folder -> asset folder. Also, copy the 2 files
-> “retrained_labels.txt” and “rounded_graph.pb”.
- The last step is to copy the TensorFlow image classifier. Also, copy the interface classifier in the same folder where is the classifier activity is.
The Necessary Code
Let us discuss the necessary code by initialization of Tensorflow.
Put in the classifier activity class inside the code -> “on create method”. Also, this method is called as “initTensorFlow and load model”.
Moreover, at the origin of the app, the TensorFlow gets initialized.
This kind of method runs only as a thread. Therefore, you should put the code at the top of this class-> “public executor executer= Executers.newSingletThreadExecutor()”.
Furthermore, you need to add a TensorFlow image classifier and classifier interface. Don’t worry it easy to understand.
You can just copy the two classes.
- Tensorflow Image Classifier
- Classifier
So right now we need to analyze a picture. This method is called the method of recognizing an image.
Firstly, you need to pick a photo on the gallery app so that it can analyze. It is a simple process.
Because it is provided from the android to the selected photo out of the gallery.
Therefore, all codes have been implemented apart. So, to interact with the codes we need to build the user interface or UI.
So basically, to add the UI we only need is to perform the event. Moreover, the image view and text view will show the result.
This is it, you can test your app and summarize it! Create your image recognition application with android.
