Apple commits 'Turi Create' machine learning development tool to GitHub

Posted:
in General Discussion
Building on its acquisition of machine learning and AI specialist Turi, Apple this week committed "Turi Create" to GitHub, a new machine learning framework designed to help developers build machine learning models that can be parlayed into apps running on its major operating systems.




According to Apple, Turi Create is designed to assist in the development and deployment of machine models, even for those not well-versed in the discipline.

With the commit, developers can add recommendations, object detection, image classification, image similarity or activity classification to their app, among other assets, Apple says.

Turi Create was posted on Thursday by Apple engineers including Gustav Larsson, Srikrishna Sridhar and Zach Nation, all of whom work on the company's CoreML machine learning framework. Sridhar and Nation were previously employees at Turi. A fourth member of the commit team, "afranklin," is also thought to be a software engineer at Apple.

Apple describes Turi Create as easy to use, with a focus on tasks instead of algorithms. The tool supports large datasets of text, images, audio, video and sensor data on a single machine, while streaming visualization help developers explore data.

Importantly, models created using Turi Create can be exported to CoreML for application in Apple's major platforms -- iOS, macOS, watchOS and tvOS.

In its ReadMe, Apple presents an example in which an image classifier is built using only a few lines of code. Accompanying images showed the resulting model, which can be integrated into an iOS app, successfully identify a flowerpot, fountain and banana.

Apple says the framework can be used for create recommender systems, image classification, image similarity, object detection, activity classifier and text classifier operations. Developers can also work with machine learning models through toolkits covering classifiers, regression, graph analytics, clustering, nearest neighbors and topic models.

Apple purchased Turi last year for a reported $200 million as it built out its machine learning and artificial intelligence team. The company's first significant effort in the space arrived with iOS 11 this year in the form of CoreML, a toolset designed to assist developers in the deployment of machine learning features.

Comments

  • Reply 1 of 9
    Nifty stuff. I'm just starting to learn about ML and see the similarities here. 
    watto_cobra
  • Reply 2 of 9
    "Turi" being short for "Turing" no doubt?
    fastasleepAirunJae
  • Reply 3 of 9
    wizard69wizard69 Posts: 12,506member
    Great now if Apple would return my MBP!


    seriously if this takes the grunt work out of machine learning solutions it will certainly help Apple solidify its position in the field.   One of the big obsticales companies have with respect to adopting this tech is the high cost of entry.  
  • Reply 4 of 9
    asdasdasdasd Posts: 5,080member
    Banana. Or not banana. 
  • Reply 5 of 9
    "Turi" being short for "Turing" no doubt?
    Turi is Siri’s younger sister.
    waverboyAirunJae
  • Reply 6 of 9
    roakeroake Posts: 563member
    "Turi" being short for "Turing" no doubt?
    Turi is Siri’s younger sister.
    Yes, Siri is the older sibling.  Older, but not so savvy, perhaps dropped on her head at birth, can’t quite understand us much of the time, but still funny and very much beloved by all.
  • Reply 7 of 9
    I hope Apple have something in the works to rival DeepMind. What that thing just did to Stockfish is very impressive, irrespective of hardware imbalance or version numbers. Google could run away with this one. 
  • Reply 8 of 9
    nhtnht Posts: 4,177member
    I hope Apple have something in the works to rival DeepMind. What that thing just did to Stockfish is very impressive, irrespective of hardware imbalance or version numbers. Google could run away with this one. 
    MSR (Microsoft Research) has almost always had better fundamental tech than Apple.  Never seemed to translate into better products.

    OpenAI is pretty cool too.  I don't think that Google is going to "run away" against all the groups working in this area...and given you can bring in TensorFlow under CoreML it doesn't matter to Apple how good TensorFlow models are.

    The three aspects of ML is algorithms, compute and data.  Google has all three but the algorithms are largely open source (except for reinforcement learning but there are others available).  Apple has the same or better mobile compute power platform.  The difference between giants is training data because they all have smart people working on core algorithms.
  • Reply 9 of 9

    In addition to its machine-learning products, Turi also runs the Data Science Summit — a two-day conference focusing on, as the name implies, data science.

    Turi was previously known as “Dato” (and before that, “GraphLab”), but changed its name in July of this year after a trademark dispute.

    Turi began reaching out to its customers to let them know their products would no longer be available at the end of July, the first indication that an acquisition had occurred. Turi’s own blog, meanwhile, no longer loads.

    Source: https://techcrunch.com/2016/08/05/apple-acquires-turi-a-machine-learning-company/

Sign In or Register to comment.