Apple‘s Create ML Empowers Developers to Harness Machine Learning

Machine learning has become a fundamental building block for the innovative apps and services that are transforming industries and reshaping our world. Tech giants like Google, Facebook, and Apple are heavily investing in ML to power more intelligent user experiences. However, machine learning has traditionally been a complex discipline requiring specialized skills, making it challenging for many developers to adopt.

Apple‘s Create ML framework, launched in 2018, aims to democratize machine learning for the vast community of developers building apps for Apple‘s ecosystem. By providing a simple, intuitive interface for training custom ML models, Create ML empowers developers to leverage the power of machine learning without needing to be experts in the underlying algorithms and techniques.

Accelerating ML Workflows with Create ML

At the core of Create ML is a suite of tools and templates that streamline the process of training machine learning models on Mac computers. Developers can use familiar tools like Xcode and Swift Playgrounds to access Create ML‘s visual interface, which supports drag-and-drop gestures for importing training data, selecting model architectures, and configuring training parameters.

Create ML offers several advantages that make it an appealing choice for iOS and macOS developers looking to integrate machine learning into their apps:

1. Faster training performance: By leveraging the GPU acceleration on Macs, Create ML enables significantly faster model training compared to traditional methods. At WWDC 2019, Apple showcased performance benchmarks indicating that Create ML could train a model to recognize yoga poses in just 29 seconds, a task that previously took 18 minutes using Turi Create and 90 minutes using Keras.

2. Model size optimization: Create ML applies optimization techniques to reduce the size of trained models without sacrificing accuracy. This is critical for mobile deployments where app package sizes are limited. In a real-world example, language learning app Memrise used Create ML to shrink their model from 90MB to just 3MB.

3. Flexible model architectures: Create ML supports training a variety of model types including neural networks, tree ensembles, support vector machines, generalized linear models, and more. This flexibility allows developers to choose the right architecture for their specific use case, whether it‘s analyzing images, text, audio, video, tabular data, or recommendations.

4. Seamless integration with Core ML: Models trained using Create ML can be easily converted to the Core ML format for integration into iOS and macOS apps. Core ML is Apple‘s framework for performing efficient on-device inference, and it has also seen significant speed and size optimizations. Core ML 3, released in 2019, delivered up to 60% faster inference and up to 75% smaller models through techniques like quantization and pruning.

Real-World Impact: Companies Succeeding with Create ML

Since its launch, Create ML has been used by numerous companies to bring the power of machine learning to their apps. At WWDC 2019, Apple invited several developers on stage to share their experiences and successes with Create ML. Some notable examples include:

• IBM Watson Services for Core ML: IBM released a set of pre-trained models for image classification, text classification, and object detection that can be easily integrated into apps using Watson Services and Core ML. Using Create ML and Swift, a developer was able to build a visual recognition app to identify different types of automobiles in just 30 minutes.

• Humane: AI startup Humane used Create ML to develop a model that detects human skeletons and estimates 3D body poses from 2D videos in real-time. Leveraging Create ML‘s transfer learning capabilities, the Humane team was able to train a highly accurate model with just 30 custom images by starting from an existing model in the Create ML library.

• Olay (Procter & Gamble): Skincare brand Olay used Create ML to power an intelligent skin analysis feature in their iOS app. The feature uses models trained on 70,000 images to identify skin concerns like wrinkles, texture, and hyperpigmentation. By using Create ML to train and optimize the models, Olay reported a 10x improvement in model training time.

These examples demonstrate the tangible impact that Create ML is having on businesses ability to leverage machine learning. By making ML more accessible, Create ML is helping a wide range of companies build smarter, more engaging, and more personalized user experiences.

The Road Ahead: Create ML‘s Evolution and Potential

As powerful of a tool as Create ML already is, Apple continues to invest in expanding its capabilities with each new release. At WWDC 2021, Apple introduced new Create ML templates for hand/body pose estimation, object detection, sound classification, and optical flow. The update also included enhancements to text classifier training like support for on-device transfer learning and data augmentation.

Looking ahead, there are several areas where we can expect Create ML to continue evolving:

1. AutoML capabilities: As the field of AutoML advances, we may see Create ML incorporate more automation in tasks like model architecture search, hyperparameter optimization, and feature engineering. This would make it even easier for developers to create high-performing models without needing to manually tweak configurations.

2. Explainable AI: There is growing interest in techniques to make ML models more interpretable and transparent. Create ML could potentially introduce features to help developers understand how models are making predictions, identify potential biases, and provide explanations to end-users.

3. Federated learning: Federated learning is an approach that allows ML models to be trained on decentralized data across multiple devices, without needing to centralize the data in the cloud. As privacy becomes an increasingly important consideration, Create ML may explore federated learning capabilities to help developers build privacy-preserving ML features.

4. Integration with other Apple services: We can expect to see deeper integrations between Create ML and other Apple AI/ML offerings like SiriKit, Core ML, Vision, and NLP. A more unified and interoperable ecosystem of tools would make it even easier for developers to infuse intelligence across the entire user experience.

As Apple continues to advance Create ML, it has the potential to be a significant driver of machine learning adoption across the Apple developer community. By democratizing access to powerful ML tools, Create ML empowers a broader set of developers to create intelligent features that can transform industries, boost productivity, and enrich people‘s lives.

Create ML and the Future of Accessible Machine Learning

Create ML represents an important step forward in making machine learning more accessible and approachable for mobile and desktop developers. It abstracts away much of the low-level complexity of training models, allowing developers to focus on designing great user experiences powered by ML.

At the same time, it‘s important to recognize that Create ML is not a replacement for the powerful, general-purpose ML platforms like TensorFlow or PyTorch. Its purpose is to make common ML tasks easier for Apple developers, not to address every possible ML use case. For more complex or specialized scenarios, developers will still need to leverage other tools and frameworks.

As AI continues to become a core differentiator in software products, we can expect to see more tech giants and startups alike investing in developer-friendly ML tools and services. Google, for example, offers tools like AutoML and ML Kit to simplify ML development within its ecosystem. The race is on to capture the hearts and minds of developers by making it as easy as possible to build intelligent apps.

In this rapidly evolving landscape, Apple is well-positioned to be a leader in accessible ML tools for developers. With Create ML, Apple has shown its commitment to empowering its developer community with easy-to-use but powerful machine learning capabilities. As Create ML continues to evolve and mature, it has the potential to accelerate the adoption of ML across the Apple ecosystem and help bring to life a new generation of smart, personalized, and deeply engaging applications.