\n The ONNX Runtime<\/a> is available across a large variety of platforms, and provides developers with the tools to run machine learning models locally. Pre-trained models can be exported<\/a> to ONNX for distribution on a variety of platforms including Android. The ONNX community<\/a> includes many companies embedding in their products, providing models, and contributing to the ecosystem.\n<\/p>\n \n To see a machine learning model in action, follow the instructions for the mobile image recognition example on Android<\/a> on the ONNX Runtime website. The example uses an existing pre-trained model \u2013 MOBILENET v2<\/a> \u2013 which is exported from ONNX format for use in the app. While there are instructions provided<\/a> to do this yourself it\u2019s easier to download<\/a> them directly if you just want to see the sample run.\n<\/p>\n \n Once you have the ONNX model and classification files, clone the Android source code<\/a> from GitHub.\n<\/p>\n \n Before you build, copy the ONNX files to the \n Note that the download<\/a> includes an older version and<\/em> a v5 version of the models \u2013 I found that I needed the v5 files (which I renamed to match the resource names in code).\n<\/p>\n \n When you run the sample and grant camera permission, the model will display the top three best matches for the subject in the video frame:\n<\/p>\n \n Figure 2: Inference percentages when detecting a dog breed<\/em>\n<\/p>\n \n This ONNX Android sample only performs image classification, but other mobile samples also cover basic usage, speech recognition, object detection, and vision-related models. Using these samples as a foundation you can incorporate other trained models<\/a> into your apps!\n<\/p>\n \n More information about the ONNX Runtime is available at onnxruntime.ai<\/a> and also on YouTube<\/a>.\n<\/p>\n \n If you have any questions about applying machine learning, or would like to tell us about your apps, use the\u00a0feedback forum<\/a>\u00a0or message us on\u00a0Twitter @surfaceduodev<\/a>.\n<\/p>\nGet started with machine Learning on Android<\/h2>\n
raw<\/code> resources directory and ensure the filenames match the Java resources references (MainActivity.java<\/strong> lines 144-154):\n<\/p>\n\/\/ Read MobileNet V2 classification labels\r\nprivate fun readLabels(): List<String> {\r\n return resources.openRawResource(R.raw.imagenet_classes).bufferedReader().readLines()\r\n}\r\n\/\/ Read ort model into a ByteArray, run in background\r\nprivate suspend fun readModel(): ByteArray = withContext(Dispatchers.IO) {\r\n val modelID =\r\n if (enableQuantizedModel) R.raw.mobilenetv2_uint8 else R.raw.mobilenetv2_float\r\n resources.openRawResource(modelID).readBytes()\r\n}<\/pre>\n![\"Screenshot](\"https:\/\/devblogs.microsoft.com\/surface-duo\/wp-content\/uploads\/sites\/53\/2023\/02\/screenshot-showing-an-example-classification-of-a.jpeg\")
\n<\/p>\nFeedback and resources<\/h2>\n