This project: includes an Android mobile application using TensorFlow Lite model trained with Stanford University Dogs Dataset consisting of 120 dog breeds and a total of 20580 images. In this application, the breed of any dog shown to the device using the camera of the mobile device are presented to the user with the similarity rate to the breed as well as the similarity rates to the other breeds. The user can also view and change how long the dog breed has been extracted takes by the application and how many threads occurs. At the same time, the processor load of the application can be changed between CPU and GPU from the user interface.
This application: aims to make people’s lives easier and gives people more information about dog breeds with machine learning and image processing algorithms. In the advanced stages of the application development, it is foreseen to provide more detailed information about dog breeds and to inform users more on this topic..
Used Items:
- TensorFlow 2.3.1
- NumPy 1.18.5
- MatPlotLib 3.2.2
- Anaconda Navigator 1.9.12
- Conda 4.8.3
- Python 3.8.3
- JupyterLab 2.1.5
- Android Studio 4.2
- Android 9.0 (Pie)
- Android SDK Platform-Tools 30.0.05
- NDK 22.0.7026061
- Pixel 2 API 28 Virtual Device
- LG-H850TR Android 8.0.0
- MobileNet V2
Here is my TFLite .ipynb file!
Here is my APK file!..
Here is my Personal Website!..
Here is Stanford Dogs Dataset!..
Here is TFLite Android Guide!..
Here is Android Studio Setup File!..
Here is Anaconda Navigator!..
Screenshot of the Application in the Test Phase of Pixel 2 API 28 Virtual Device
Screenshot of the Application in the Test Phase of LG-H850TR Device
Screenshot of Sub-Menu in the Test Phase of LG-H850TR Device
NOTE: This Project is "Computer Engineering Project Design" Lesson Work by Mahmut Can Kurt.
Here is my PROJECT REPORT (TR)!..