Skip to content

Anku5hk/Android-Model-Tester

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

About

This Android application is used to test inferences of Tensorflow/Pytorch graphs. This application supports variety of Models, Computer Vision tasks independent of library. Also one can use it to learn how the code works for pytorch and tensorflow. Soon more tasks will be added gradually.

Features

  • Fast inference on android for tensorflow/pytorch graphs.
  • Supports Classification, Object Detection(Faster R-CNN).
  • Simple GUI based options for inputs.
  • Auto detect library(Tensorflow/Pytorch).
  • Small size and easy to understand code(Tons of comments).
  • Open source: Edit/Make changes, use as a starter template No Worries.

How To?

  1. Train your model.
  2. Save it as .pt for pytorch or .tflite for tensorflow.
  3. Put labels(classes names) into a simple .txt file, no formatting needed only newline for new class. Same for object detection.
  4. Select files through application setting.
  5. Select image and press button.
  6. Done.

Screens

3 1 2 1

More Features to be added.

  • More Image tasks.

More

You can compile above code into Android studio or download apk
1.0 here, Update 1.1 here
Classification Models Zoo:
Pytorch (https://github.com/qubvel/classification_models)
Tensorflow (https://www.tensorflow.org/lite/guide/hosted_models)
Object Detection Models Zoo: (https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md)

Update

  • Changed code to completely follow the official documentations.
  • Added support for camerax.
  • Tensorflow runs on 0.0.0-nightly version.
  • Supports Quant/Float models on Tensorflow.

About

Android app to test Tensorflow/Pytorch graphs.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published