Every developer loves TensorFlow and even more when you can implement it directly on Android. But we have to do a lot before that, right? Suppose we’re performing an image classification task. In that case, you’ll probably get a Bitmap or a new AI-based system Image object from the Camera library, and then we transform it into a float[][][] or a byte[]. This is not even the last step, we need to do a lot more, and there is no other way around.
But TensorFlow has introduced a new library to solve the tedious tasks of pre-processing, TensorFlow Lite Task Library.
TensorFlow Lite is presently in developer preview, so it may not support all TensorFlow models’ operations. Despite this, it does work with standard Image Classification models, including Inception and MobileNets.
TensorFlow Lite Task Library is specifically designed to achieve the best performance and usability. The traditional tedious ways of image classification, object detection, and other machine learning tasks can now be done with just five lines of codes!
TensorFlow Lite is widely implemented on various apps like pose estimation, speech recognition, gesture recognition, smart reply, style transfer, on-device recommendation, and many more.Â
Source: https://blog.tensorflow.org/2020/09/introducing-tensorflow-lite-task-library.html
Consulting Intern: Grounded and solution--oriented Computer Engineering student with a wide variety of learning experiences. Passionate about learning new technologies and implementing it at the same time.
