What is Core ML from Coding
What is Core ML – Apple Machine Learning Framework
With Core ML it is possible to integrate well-trained machine-learning models into your own iOS apps. It’s an Apple-provided framework that’s optimized for machine learning applications on Apple devices, minimizing storage and power consumption.
[Related Article: Artificial Intelligence Trends ]
The framework is suitable for various Apple products and works together, for example, with
[Related Article: Python For Machine Learning ]
Other supported frameworks include MXNet or TensorFlow . AI applications can be used for voice and image recognition, for text extraction, or for finding relationships and patterns in data.
[Related Article: Machine Learning vs. Deep Learning]
The main features of Core ML
Core ML supports a variety of different models. For maximum performance, Core ML uses the power of CPUs and GPUs. The machine learning models and applications are fully operational on a dedicated device without the need to transfer data to other systems for analysis. This ensures the protection of the data used and allows the application to function reliably even without a network.
Core ML supports vision for image analysis and Natural Language for text analysis. Vision’s possible machine-learning features include facial recognition, text recognition, barcode recognition, object tracking, person tracking, and more.
[Related Article: What is a Convolutional Neural Network?]
Use of converters to integrate third-party machine learning models with Core ML
Core ML is not intended to create and train machine learning models. The framework relies on already created and trained models. Apple provides some sample models and Create ML to create and train custom models.
[Related Article: The Deep Learning Framework]
In addition, there are numerous converters with which machine learning models of various frameworks and formats can be converted into the format required by Core ML. For example, converters are available for the following frameworks:
- Apache MXNet
- TensorFlow
- ONNX (Open Neural Network Exchange)
[Related Article: What is Cognitive Computing? ]
The interplay of Create ML and Core ML
In addition to Core ML, Create ML is an important tool from Apple to implement its own machine learning applications. Core ML was introduced at the Apple Developers Conference WWDC 2017. Create ML followed a year later. While Core ML integrates ready-made model into its own apps, Create ML is designed for creating and training models.
Create ML uses Apple technologies such as Swift or Xcode and is able to automate the creation and training of models using Swift scripts. The models can be tested and trained on a Mac computer and do not need dedicated servers. Data can be fed via drag and drop via a graphical user interface. Developers do not need specific knowledge about the algorithms used or the structure of the underlying neural network.
[Related Article: Artificial Intelligence And The Blockchain ]
Benefits of using Core ML and Create ML
Core ML and Create ML allow developers to incorporate artificial intelligence and machine learning functions into their own apps with little effort and without special programming skills. The frameworks are easy to use and work on Apple devices with high performance.
[Related Article: Amazon Machine Learning ]
Training is possible directly on local computers and does not require its own server infrastructure. The finished apps work standalone on Apple devices. No data needs to be transferred to external devices and no internet connection is required. Services or software from external providers are superfluous. If necessary, models of other frameworks such as TensorFlow can be adopted via Core ML.
[Related Article: Artificial Intelligence Vs Augmented Intelligence ]
Improvements in the current version of Core ML 2
The current version of Core ML is Core ML 2 (as of March 2019). It was presented at WWDC 2018 by Apple. Core ML 2 delivers even more performance and enables the use of even more compact machine learning models on iOS devices. The new version allows developers to access a wide range of machine learning models. This includes standard models such as SVM (Support Vector Machines) or Tree Ensembles and more than 30 Deep Learning Layer types.