Deep Learning on iOS

At Trifork Machine Learning, we pride ourselves in working the latest developments in Machine Learning. Recent advancements in the field of Machine Learning have been particularly focused on Deep Learning. If we look at the below trend, it has certainly experienced a surge in attention of the last couple of years. News in the form of self-driving cars and speech robots permeates our everyday lives, but how can we use Deep Learning to improve businesses?

https://trifork.com/wp-content/uploads/2018/11/Deep-Learning-trend.jpg

Deep Learning

Deep Learning has enabled its practitioners to effectively make use of previously unusable (unstructured) data sources such as texts, images and video or audio streams. While these models are exceptionally good at extracting high-level patterns from very large sets of data, they tend to require a lot in terms of computing power. Deep Learning models are therefore typically constructed using dedicated AI chips, such as a GPU or TPU, rather than traditional CPUs. Deep Learning models take long to train and the parallelisation options in a GPU help with keeping this in check.

On the Edge

Typically, Deep Learning models would be hosted in a scalable environment (such as AWS or GCP), but this is not always desired. An alternative is to run the trained model on a mobile device. One of our customers required object detection tasks on mobile devices because of limitations in connectivity. Furthermore, the images taken by its users contain sensitive information, and should therefore not be sent back and forth to a server. Despite the demands in computing power, it is quite possible to have Deep Learning models run directly on mobile devices, or more specifically in this case, iOS devices.

https://trifork.com/wp-content/uploads/2018/11/Web-shutterstock_780954301-2.jpg
https://trifork.com/wp-content/uploads/2018/11/DeepLearning.png

CoreML

The latest developments of Apple’s CoreML solution for iOS allowed us to transfer a Deep Learning model to iPhone with relative ease. In addition, CoreML allows you to use the iPhone’s built-in dedicated AI chip out of the box, which helps with the required performance.
Do you want to know more about this project? Contact us.

Contact us

Would you like to know more about optimising your business with Deep Learning solutions on iOS? Please fill out the contact form.

 

Privacy
Trifork always treats your personal details with the utmost respect and use them to respond to all your requests. Please notify us if you wish to be excluded from our database.