{"id":1399,"date":"2017-07-07T11:11:40","date_gmt":"2017-07-07T11:11:40","guid":{"rendered":"http:\/\/www.tooploox.com\/?p=1399"},"modified":"2024-04-15T13:56:25","modified_gmt":"2024-04-15T13:56:25","slug":"custom-classifiers-in-ios11-using-coreml-and-vision","status":"publish","type":"post","link":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision","title":{"rendered":"Custom classifiers in iOS11 using CoreML and Vision"},"content":{"rendered":"\r\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png\" alt=\"\"\/><\/figure>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Introduction<\/h2>\r\n\r\n\r\n\r\n<p>The hype for Machine Learning, Deep Learning and AI is in its peak. Right in the middle of that, Apple provided developer community with tools to run trained models on its hardware. It&nbsp;is a pretty big deal as earlier there was no way for engineers to easily deploy such applications on multiple physical devices. We thought it would be nice to have our turn with the&nbsp;<em>CoreML<\/em>&nbsp;and so the iOS and Data Science decided to team up. Therefore we decided to go the extra mile and present our full pipeline from model conversion using&nbsp;<em>coremltools<\/em>&nbsp;python package to actual iOS implementation. Let me guide you through our journey. To follow, you are going to need Xcode 9, device with beta of iOS11 installed (simulator won\u2019t do as camera has to be accessed) and python 2.7 installed on your Mac (it should be by default just as&nbsp;<em>python<\/em>).<\/p>\r\n\r\n\r\n\r\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><img decoding=\"async\" src=\"https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/IMG_4587.png\" alt=\"\"\/><\/figure><\/div>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Caffe and Python introduction<\/h2>\r\n\r\n\r\n\r\n<p>First we have to find and convert trained model to&nbsp;<em>.mlmodel<\/em>&nbsp;format. We got ours from&nbsp;<a href=\"https:\/\/github.com\/BVLC\/caffe\/wiki\/Model-Zoo#models-for-age-and-gender-classification\">Model Zoo<\/a>&nbsp;repository, precisely from&nbsp;<a href=\"https:\/\/talhassner.github.io\/home\/publication\/2015_CVPR\">this<\/a>&nbsp;project. Scroll to downloads section where you can find&nbsp;<em>cnn_age_gender_models_and_data.0.0.2.zip<\/em>&nbsp;&nbsp;file with everything we used here (about 86 MB download). Inside, there are 2 trained classifiers and some additional files. Model obtained from&nbsp;<em>Caffe<\/em>&nbsp;deep learning framework comes in 2 files:&nbsp;<em>.prototxt<\/em>&nbsp;and &nbsp;<em>.caffemodel<\/em>. The former one encodes an architecture for the neural network while the latter contains learned weights for a network connection. For further work we picked the Age Net (<em>age_net.caffemodel<\/em>,&nbsp;<em>deploy_age.prototxt<\/em>) and Gender Net (<em>gender_net.caffemodel<\/em>,&nbsp;<em>deploy_gender.prototxt<\/em>).<\/p>\r\n\r\n\r\n\r\n<p>Let\u2019s get our hands dirty. I won\u2019t dwell too much on python environment details and converted models are included in the&nbsp;<a href=\"https:\/\/github.com\/jknthn\/core-ml-demo\">repository<\/a>. I feel much safer using virtual environment for any python development. As far as I am concerned, anaconda is worth recommending. You can read more about this concept here:<\/p>\r\n\r\n\r\n\r\n<p>\u2013&nbsp;<a href=\"https:\/\/python-guide-pt-br.readthedocs.io\/pt_BR\/latest\/dev\/virtualenvs.html\">virtualenv<\/a><\/p>\r\n\r\n\r\n\r\n<p>\u2013&nbsp;<a href=\"https:\/\/continuum.graphics\/downloads\/\">anaconda<\/a><\/p>\r\n\r\n\r\n\r\n<p>Having our virtual environment set up (remember python 2.7 for\u00a0<em>coremltools<\/em>, 3.0+ is not supported, been there, done that). Go to terminal and install\u00a0<em>coremltools<\/em>\u00a0and\u00a0<em>jupyter-notebook<\/em>\u00a0to code it in.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/271616040b6262700a36e9b6ded01c75.js\"><\/script>\r\n\r\n\r\n\r\n<p>The two first commands install both of the needed packages using\u00a0<em>pip<\/em>\u00a0python package manager. Third runs\u00a0<em>jupyter-notebook<\/em>\u00a0environment which should open in your browser. If you are working inside a virtual environment, you should also add your local python kernel to available python kernels.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/de5bbf5293951005c81d984f025848fb.js\"><\/script>\r\n\r\n\r\n\r\n<p>Now you are ready to create a new notebook using appropriate python version. Again, I don\u2019t want to delve into details to much, as it is not the subject of this post. If you have any questions regarding python or jupyter environment, googling it out will probably yield pretty decent answers and you could always ask here.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Converting to&nbsp;<em>.mlmodel<\/em><\/h2>\r\n\r\n\r\n\r\n<p>After setting up everything with python the rest is pretty straightforward.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/c665c5d253dc6d8e460ac8061fbdb894.js\"><\/script>\r\n\r\n\r\n\r\n<p>It is the simplest and fastest implementation. We are not messing up with all the labels to make it looking beautiful in Xcode, but feel free to have fun on your own. The process is similar for other models. For example in\u00a0<em>Keras<\/em>\u00a0you have\u00a0<em>.h5<\/em>\u00a0file defining the network, you just call\u00a0<em>coremltools.converters.keras.convert<\/em>\u00a0to proceed with some optional parameters like\u00a0<em>image_input_names<\/em>\u00a0and\u00a0<em>class_labels<\/em>. Let\u2019s look at\u00a0<em>deploy_gender.prototxt<\/em>\u00a0file.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/d2d10ae17c80488379951c2f9e0cc880.js\"><\/script>\r\n\r\n\r\n\r\n<p>This fragment defines the input of a neural network. I encourage you to take a look at .<em>prototxt<\/em>\u00a0files, especially if you are familiar with neural networks. So we have an input called \u201cdata\u201d and it\u2019s sized 1x3x277x277. Without specyfing it as\u00a0<em>image_input_names<\/em>\u00a0we would have to deal with\u00a0<a href=\"https:\/\/developer.apple.com\/documentation\/coreml\/mlmultiarray\">MLMultiArray<\/a>, which is just multi-dimensional array implementation. As you know, we could pass an image to this classifier passing\u00a0<em>image_input_names=\u2019data\u2019<\/em>\u00a0to the\u00a0<em>convert<\/em>\u00a0function proper input type. As for the\u00a0<em>class_labels<\/em>, they are are basically plain text files, which contain names for prediction results separated by newlines.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/0faba86e2a9f8fd136b4ad0c15279229.js\"><\/script>\r\n\r\n\r\n\r\n<p>Here I created labels for age classifier taken from model documentation. Raw classifier returns results as&nbsp;<em>MLMultiArray<\/em>&nbsp;of probabilities and we take max value as its prediction. We need&nbsp;<em>age_labels.txt<\/em>&nbsp;so our model can later output just what we need \u2013 readable prediction instead of probability numbers. Of course, you can get the numbers too if you need them. You can find all the files in the repository. After using&nbsp;<em>save<\/em>&nbsp;method on the model, it will create&nbsp;<em>.mlmodel<\/em>&nbsp;files in your script working directory. Next part is as easy as Apple promised- just drag&nbsp;<em>.mlmodels<\/em>&nbsp;to your Xcode project.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Using the model<\/h2>\r\n\r\n\r\n\r\n<p>What happens after you import model to Xcode? First and foremost, Swift code for interfacing with freshly created model. At this point you have no power to modify it anymore. All the parameters, even meta ones like author or various descriptions, have to be set up via&nbsp;<em>coremltools<\/em>. I imagine it\u2019s a way for the original model authors to be attributed and mentioned but the current ecosystem is full of open-sourced models, which you have to convert yourself. Also, all generated class names are taken from .<em>mlmodel<\/em>&nbsp;file name so if you want your code to be nice and consistent, do name it in PascalCase.<\/p>\r\n\r\n\r\n\r\n<p>What about the code which is generated for us? We\u2019ve got 3 classes:\u00a0<em>YourModelInput<\/em>,\u00a0<em>YourModelOutput<\/em>\u00a0and\u00a0<em>YourModel<\/em>. Input and output are here for the things to be very swifty and safe, but those are more or less wrappers on the intended values. In our case input for both models is an Image of size 277 x 277. Unfortunately, it\u2019s not\u00a0<em>UIImage<\/em>\u00a0or\u00a0<em>CGImage<\/em>\u00a0but\u00a0<em>CVImageBuffer<\/em>\u00a0which is raw pixel data (or to be more specific a pointer to it). Bad thing about it- we don\u2019t have any tools, neither Apple\u2019s nor open source to manipulate it in a good way. Our initial solution was to convert buffer to\u00a0<em>UIImage<\/em>, manipulate and then convert it back to buffer. And it works. There is probably performance bottleneck here, but it was good enough as a quick and dirty solution. The output is an array of type\u00a0<em>MLMultiArray<\/em>\u00a0which contains probabilities for each predicted class and if we provided python script with class labels file, it also contains label of most probable class according to the predictive algorithm. Model class is used to make prediction. You can find all helper code in our repository.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/8f339d87cad40cc1dc1d3dfbaa7f3551.js\"><\/script>\r\n\r\n\r\n\r\n<p>Unfortunately, the described solution has its disadvantages, especially when it comes to a real time video processing. For each video frame we have to convert image to&nbsp;<em>UIImage<\/em>, scale and then back to&nbsp;<em>CVPixelBuffer<\/em>. Our modern hardware can handle it without too much fps loss of preview, but I consider it a waste and a really bad practice.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Enter the&nbsp;<em>Vision<\/em><\/h2>\r\n\r\n\r\n\r\n<p>The new\u00a0<em>Vision<\/em>\u00a0API introduced recently, besides all marketed features, also allows to make predictions with your trained machine learning models. And you can pass either\u00a0<em>CGImage<\/em>\u00a0or\u00a0<em>CVPixellBuffer<\/em>\u00a0to it, which is nice. No additional conversion back to buffer is required. There are 2 main classes you need to know about.\u00a0<em>VNCoreMLRequest<\/em>, which you initialize with the model and completion closure and\u00a0<em>VNImageRequestHandler<\/em>\u00a0containing in our case\u00a0<em>CVImageBuffer<\/em>. Making prediction is as simple as calling\u00a0<em>perform<\/em>\u00a0method on\u00a0<em>VNImageRequestHandler<\/em>\u00a0object with array of \u00a0<em>VNCoreMLRequest<\/em>\u00a0as a parameter. Voila! No need for scaling and converting images over and over. Well played Apple, well played. The neat thing is that you can make many predictions in one function call, as in our case, to classify age and gender at once. Let me give you a little snippet and go to the repository for full code.<\/p>\r\n\r\n\r\n\r\n<script src=\"https:\/\/gist.github.com\/jknthn\/f0068907a46659501477b76a62a3cb19.js\"><\/script>\r\n\r\n\r\n\r\n<p>The prediction is returned in the form of a tuple&nbsp;<em>(VNRequest, Error?)<\/em>&nbsp;and you can get an array of probabilities using&nbsp;<em>VNRequest<\/em>&nbsp;<em>results<\/em>&nbsp;property. They are sorted descendingly, so first element in an array is our prediction result wrapped in&nbsp;<em>VNClassificationObservation<\/em>&nbsp;class. Two properties are worth noting here.&nbsp;<em>Identifier<\/em>&nbsp;contains label and its confidence` probability. From here you can update your UI, but just remember you do it on main thread when using live camera feed, as we did here. It isn\u2019t obvious, but camera and its feed work in a background thread. The above-described solution is little workaround to make one prediction for 2 classifiers work using&nbsp;<em>VNImageRequestHandler<\/em>. I didn\u2019t manage to get into internals of&nbsp;<em>Vision<\/em>. Documentation is scarce at this moment, but I imagine (it might be a bug, we are still deep in beta) there is some state sharing between request. Requests created once as property did not work, same with different closures when initializing them. Current solution looks very quick and dirty, so if you find a better way to deal with it, please let me know.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Repository<\/h2>\r\n\r\n\r\n\r\n<p>All application code with helper extension and class you can find&nbsp;<a href=\"https:\/\/github.com\/jknthn\/core-ml-demo\">here<\/a>.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Demo<\/h2>\r\n\r\n\r\n\r\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><img decoding=\"async\" src=\"https:\/\/tooploox.com\/wp-content\/uploads\/2023\/02\/3_giphy.gif\" alt=\"\"\/><\/figure><\/div>\r\n\r\n\r\n\r\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><img decoding=\"async\" src=\"https:\/\/tooploox.com\/wp-content\/uploads\/2023\/02\/4_giphy.gif\" alt=\"\"\/><\/figure><\/div>\r\n\r\n\r\n\r\n<p><strong>Read also about&nbsp;<a href=\"https:\/\/tooploox.com\/apple-app-clips\" target=\"_blank\" rel=\"noreferrer noopener\">Apple app clips, new feature for iOS 14<\/a>.<\/strong><\/p>\r\n\r\n\r\n\r\n<p><\/p>\r\n","protected":false},"excerpt":{"rendered":"<p>Introduction The hype for Machine Learning, Deep Learning and AI is in its peak. Right in the middle of that, Apple provided developer community with tools to run trained models on its hardware. It&nbsp;is a pretty big deal as earlier there was no way for engineers to easily deploy such applications on multiple physical devices. [&hellip;]<\/p>\n","protected":false},"author":9,"featured_media":1416,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","_uag_custom_page_level_css":"","footnotes":""},"categories":[113,312],"tags":[177,191,192,236],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v21.5 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Custom classifiers in iOS11 using CoreML and Vision - Tooploox<\/title>\n<meta name=\"description\" content=\"Discover how to utilize CoreML and Vision in iOS 11 for custom classifiers, featuring step-by-step guides on converting .caffemodel files with coremltools. This article provides a practical approach to integrating advanced machine learning models into iOS applications, enhancing app capabilities with powerful image recognition.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision\" \/>\n<meta property=\"og:locale\" content=\"en_GB\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Custom classifiers in iOS11 using CoreML and Vision - Tooploox\" \/>\n<meta property=\"og:description\" content=\"Discover how to utilize CoreML and Vision in iOS 11 for custom classifiers, featuring step-by-step guides on converting .caffemodel files with coremltools. This article provides a practical approach to integrating advanced machine learning models into iOS applications, enhancing app capabilities with powerful image recognition.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision\" \/>\n<meta property=\"og:site_name\" content=\"Tooploox\" \/>\n<meta property=\"article:published_time\" content=\"2017-07-07T11:11:40+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-04-15T13:56:25+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png\" \/>\n\t<meta property=\"og:image:width\" content=\"3800\" \/>\n\t<meta property=\"og:image:height\" content=\"1600\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Jeremi Kaczmarczyk\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Jeremi Kaczmarczyk\" \/>\n\t<meta name=\"twitter:label2\" content=\"Estimated reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Custom classifiers in iOS11 using CoreML and Vision - Tooploox","description":"Discover how to utilize CoreML and Vision in iOS 11 for custom classifiers, featuring step-by-step guides on converting .caffemodel files with coremltools. This article provides a practical approach to integrating advanced machine learning models into iOS applications, enhancing app capabilities with powerful image recognition.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision","og_locale":"en_GB","og_type":"article","og_title":"Custom classifiers in iOS11 using CoreML and Vision - Tooploox","og_description":"Discover how to utilize CoreML and Vision in iOS 11 for custom classifiers, featuring step-by-step guides on converting .caffemodel files with coremltools. This article provides a practical approach to integrating advanced machine learning models into iOS applications, enhancing app capabilities with powerful image recognition.","og_url":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision","og_site_name":"Tooploox","article_published_time":"2017-07-07T11:11:40+00:00","article_modified_time":"2024-04-15T13:56:25+00:00","og_image":[{"width":3800,"height":1600,"url":"https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png","type":"image\/png"}],"author":"Jeremi Kaczmarczyk","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Jeremi Kaczmarczyk","Estimated reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision","url":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision","name":"Custom classifiers in iOS11 using CoreML and Vision - Tooploox","isPartOf":{"@id":"https:\/\/tooploox.com\/#website"},"datePublished":"2017-07-07T11:11:40+00:00","dateModified":"2024-04-15T13:56:25+00:00","author":{"@id":"https:\/\/tooploox.com\/#\/schema\/person\/88bc0b77b46573fb12cd99d80bf3c328"},"description":"Discover how to utilize CoreML and Vision in iOS 11 for custom classifiers, featuring step-by-step guides on converting .caffemodel files with coremltools. This article provides a practical approach to integrating advanced machine learning models into iOS applications, enhancing app capabilities with powerful image recognition.","breadcrumb":{"@id":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision#breadcrumb"},"inLanguage":"en-GB","potentialAction":[{"@type":"ReadAction","target":["https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/tooploox.com\/custom-classifiers-in-ios11-using-coreml-and-vision#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/tooploox.com\/"},{"@type":"ListItem","position":2,"name":"Custom classifiers in iOS11 using CoreML and Vision"}]},{"@type":"WebSite","@id":"https:\/\/tooploox.com\/#website","url":"https:\/\/tooploox.com\/","name":"Tooploox","description":"Tooploox helps startups and tech-innovators deliver top-notch products by applying mobile, web, and AI solutions in a comprehensive and reliable way","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/tooploox.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"en-GB"},{"@type":"Person","@id":"https:\/\/tooploox.com\/#\/schema\/person\/88bc0b77b46573fb12cd99d80bf3c328","name":"Jeremi Kaczmarczyk","image":{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/tooploox.com\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/f84a602399350d0bbd6bed53a005dbb4?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/f84a602399350d0bbd6bed53a005dbb4?s=96&d=mm&r=g","caption":"Jeremi Kaczmarczyk"}}]}},"uagb_featured_image_src":{"full":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png",3800,1600,false],"thumbnail":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-150x150.png",150,150,true],"medium":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png",300,126,false],"medium_large":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png",768,323,false],"large":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png",1024,431,false],"1536x1536":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-1536x647.png",1536,647,true],"2048x2048":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-2048x862.png",2048,862,true],"1920-auto":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-1920x808.png",1920,808,true],"1560-auto":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-1560x657.png",1560,657,true],"1400-800":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-1400x800.png",1400,800,true],"1200-900":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-1200x900.png",1200,900,true],"820-820":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-820x820.png",820,820,true],"700-auto":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-700x295.png",700,295,true],"570-auto":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-570x240.png",570,240,true],"420-auto":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-420x177.png",420,177,true],"350-500":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-350x500.png",350,500,true],"240-180":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-240x101.png",240,101,true],"240-240":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2.png",240,101,false],"post-thumbnail":["https:\/\/tooploox.com\/wp-content\/uploads\/2017\/07\/coreml2-250x250.png",250,250,true]},"uagb_author_info":{"display_name":"Jeremi Kaczmarczyk","author_link":"https:\/\/tooploox.com\/author\/jeremi-kaczmarczyk"},"uagb_comment_info":0,"uagb_excerpt":"Introduction The hype for Machine Learning, Deep Learning and AI is in its peak. Right in the middle of that, Apple provided developer community with tools to run trained models on its hardware. It&nbsp;is a pretty big deal as earlier there was no way for engineers to easily deploy such applications on multiple physical devices.&hellip;","_links":{"self":[{"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/posts\/1399"}],"collection":[{"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/comments?post=1399"}],"version-history":[{"count":1,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/posts\/1399\/revisions"}],"predecessor-version":[{"id":16750,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/posts\/1399\/revisions\/16750"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/media\/1416"}],"wp:attachment":[{"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/media?parent=1399"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/categories?post=1399"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tooploox.com\/wp-json\/wp\/v2\/tags?post=1399"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}