What is the MindSpore Lite of the HMS Core Video Editor Kit? MindSpore Lite Helps HMS Core Video Editor Kit Deliver Intelligent Video Editing Experience.

By | May 10, 2022

New media — like vlogs and short videos — are on fire, as they ride the wave of changes brought by mobile phones on how we socialize and keep entertained. Photo and video editing apps are therefore becoming more functional, utilizing AI-empowered functions for a range of situations.

Such an editing app is expected to have various functions and materials, which (logically) entails a long development period and high requirements. Creating such an app is a daunting task for developers, one that is both slow and demanding. Video Editor Kit, launched in HMS Core 6.0, helps developers build a smart, easy-to-use editing app with a range of capabilities including resource input, editing, rendering, and output, as well as material management. In addition to these common functions, the kit also offers advanced, AI-driven functions such as AI filter, track person, and color hair, allowing users to fully unleash their creativity.

Demo for Video Editor Kit’s capabilities

These powerful functions must come from somewhere, and that somewhere is neural network models. A single model can exceed 10 MB. Together, those models can occupy considerable ROM and RAM space on a device. Therefore, another challenge for video editing app developers is to ensure that their app occupies as little space as possible.

To resolve this concern, Video Editor Kit turns to MindSpore Lite (a Huawei-developed AI engine) for inference of neural network models. MindSpore Lite is a fantastic solution to this because of its inference neural network models. It is loaded with unified APIs, supporting flexible model deployment from devices to edges and cloud. It is available on devices with the Ascend processor, GPU, CPU (whose architecture can be X86, ARM, and more) or other hardware units, and can run on operating systems such as HarmonyOS, Android, iOS, Windows, or others. In addition to models trained by MindSpore, MindSpore Lite can convert and infer third-party models from TensorFlow, TensorFlow Lite, Caffe, ONNX, and more.

Demo for Video Editor Kit’s capabilities

These powerful functions must come from somewhere, and that somewhere is neural network models. A single model can exceed 10 MB. Together, those models can occupy considerable ROM and RAM space on a device. Therefore, another challenge for video editing app developers is to ensure that their app occupies as little space as possible.

To resolve this concern, Video Editor Kit turns to MindSpore Lite (a Huawei-developed AI engine) for inference of neural network models. MindSpore Lite is a fantastic solution to this because of its inference neural network models. It is loaded with unified APIs, supporting flexible model deployment from devices to edges and cloud. It is available on devices with the Ascend processor, GPU, CPU (whose architecture can be X86, ARM, and more) or other hardware units, and can run on operating systems such as HarmonyOS, Android, iOS, Windows, or others. In addition to models trained by MindSpore, MindSpore Lite can convert and infer third-party models from TensorFlow, TensorFlow Lite, Caffe, ONNX, and more.

Model quantization effect for Video Editor Kit capabilities (sourced from the test data of MindSpore)

With model quantization and compression, Video Editor Kit allows more of its AI models to be deployed in an app, without occupying more ROM space, helping equip an app with smarter editing capabilities.

Just take a look at Petal Clip, Huawei’s own dedicated editing app. After integrating Video Editor Kit, the app now offers intelligent editing functions such as AI filter and track person, and more capabilities driven by the kit will become available during later app updates. With those functions, Petal Clip makes video editing more fun and straightforward.

The high-performance, ultra-lightweight AI engine of MindSpore Lite is not just armed with powerful kernel algorithms, heterogeneous hardware scheduling, and model quantization compression, but also provides one-stop model training and inference capabilities that feature the device-cloud synergy. Together with MindSpore Lite, Video Editor Kit paves the road for developing a straightforward, intelligent editing mobile app.

To find more, please visit:

HUAWEI DevelopersHMS Core

Leave a Reply

Your email address will not be published.