Skip to content

OutoftheBoxFTC/EasyTensorflowAPI

Repository files navigation

EasyTensorflowAPI

An easy to use API to use Tensorflow 2.0 in FTC

FIRST provides a decent Tensorflow API already packaged in the app, but it currently only supports Tensorflow 1.0 and does not expose many backend features for user customizability.

In addition, due to the nature of tensorflow models, being able to see the implementation of Tensorflow allows for people to make much more specific models for FTC use

Disclaimer

As of right now, the Tensor Image Classification system is only partially tested, testing will come soon.

Current Features:

EasyTensorflowAPI currently supports both tensorflow object detection and tensorflow image classification models in the form of a .tflite model

The API supports loading models from the assets folder of FTCRobotController, setting used threads, managing NNAPI delegate usage, GPU acceleration, XNNPack, and more

Models are all run on mat inputs, interfacing seamlessly with EasyOpenCV with minimal overhead and memory leakage

Documentation:

TensorImageClassifier

TensorObjectDetector

How to make a Tensorflow Object Detection Model: WIP

How to make a Tensorflow Image Classification Model: WIP

Installing

Installation is simple. Just go to build.dependencies.gradle in your project and add the following

maven { url 'https://jitpack.io' }

in repositories and

implementation 'com.github.OutoftheBoxFTC:EasyTensorflowAPI:v1.1.0-Alpha'

in dependencies. Then run a gradle sync, and everything should download!

OnBotJava:

This library can be ported to OnBotJava, however I strongly recommend anyone using this library to use it in Android Studio to avoid a lot of pain that will come from porting it.

Important Note:

This API is still a work in progress and is subject to change. The basic usage of the API will remain the same for the forseeable future, but new features will be added as teams share feedback on things they want to see.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages