ABSTRACT

Before colonization, there were over 250 languages spoken in Australia. Today only thirteen Indigenous languages are still being taught to children). Language has an important part to play in cultural maintenance and ‘closing the gap’ in terms of First Peoples’ cultural heritage, identity, and sense of belonging. In this work, we aim to develop an engaging and easy way to teach and learn the local Indigenous names of wildflowers using a mobile device. This paper presents the development of a phone application that runs on a local machine, recognizes local wildflowers through its camera, and plays associated sounds and displays associated text in the Noongar language. The prototype mobile application has been developed with MobileNets model on the TensorFlow platform. The dataset is derived from Google searches, while the sound files are generated from label text by running an apple script. UI and interactivity have been developed by using Vuforia and the Unity game engine. Finally, the Android Studio is used to deploy the app. At this point in time, the prototype can only recognize ten local flowers, with 85%∼99% of accuracy. We are working with a larger dataset towards developing the full application.

摘要

在殖民之前，澳大利亚使用超过250种语言。如今，仍仅向儿童教授十三种土著语言。语言在维持原住民的文化遗产，身份和归属感方面在文化维护和“缩小鸿沟”中发挥着重要作用。在这项工作中，我们旨在开发一种引人入胜的简便方法，以使用移动设备教授和学习野花的当地土著名称。本文介绍了在本地计算机上运行的电话应用程序的开发，该电话应用程序通过其相机识别本地野花，并播放相关声音并以Noongar语言显示相关文本。已在TensorFlow平台上使用MobileNets模型开发了原型移动应用程序。数据集来自Google搜索，而声音文件是通过运行Apple脚本从标签文本生成的。通过使用Vuforia和Unity游戏引擎已经开发了UI和交互性。最后，使用Android Studio部署应用程序。在这个时间点上，原型只能识别10朵本地花，准确度为85％〜99％。我们正在使用更大的数据集来开发完整的应用程序。