Only a few decades ago it seemed impossible to automatically classify bird species in images, let computers translate texts from German to English in a way that doesn't insult readers, or ask voice assistants for tomorrow's weather.
All these applications were enabled by the rise of Deep Learning. When it comes to state-of-the-art results in analyzing unstructured data - the kind of which you find in images, audio, or text - deep neural networks are the way to go.
Fortunately, together with progress in research, frameworks like Tensorflow and PyTorch emerged to ease development. It is now possible to download pre-trained models from the internet, fit them to one's need or just use them.
Tensorflow Lite is an inference framework that makes it possible to run such models on embedded hardware and mobile devices. Its C++ API makes it easy to integrate use cases like computer vision or keyphrase detection into applications written with Qt. In this talk, we explain the use of Tensorflow Lite with Qt and QML, and demonstrate how everything works fluently on low-level hardware such as the Raspberry Pi.