Android camera project
This is a android application aimed to run TensorFlow image recognization modles on mobile phone, to test their working efficiency and give quick report in real time on your phone screen.
This application works better on phones with SDK API Level21(Android 5.0)
- Download project and import all files into your Android Studio
- Run dependency download
- Goto folder name assets under path app/main/src, and put your modles in it. Per requirment from Google, file name should not contain underscore.
- Go find approximately line 216, examples are below:
public modelUtils model = new modelUtils(modelUtils.ResNet50_PATH,"input_2","output_1","traditional",224);
change the things there accordingly, all 5 parameters may needs changing, you may feel free to add some constant value in modelUtils class, in which I defined something to organize model utilization.
Parameter explaination: public String MODEL_PATH; //Path of model public String INPUT_NAME; //Input layer name of model public String OUTPUT_NAME; //Output layer name of model public int Resize_Para; //Model's input img size public String OUTPUT_TYPE; //Output of model, see as MainActicity.java line 569 Class Monitor. It is advisible for you to make some change to it, to make it more handy for yourself.
- Test your app as your wish.
**Actually, currently I have added file selection function, but I am still working on customlizing parameter part so I would not release it until it is fully operational **
if get vacant, I would make it easier to adjust. like make modles more easier to remove or add.
Feel free to contact me via email: lv4399@126.com, as you can see from my account, I am a Chinese developer.