{"product_id":"coral-usb-accelerator","title":"Coral USB Accelerator","description":"\u003cp\u003e\u003cspan\u003eThe Coral USB Accelerator is a USB accessory that brings machine learning inferencing to existing systems. It works with the Raspberry Pi and Linux, Mac, and Windows systems.\u003c\/span\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eThe Accelerator \u003cstrong\u003eadds an Edge TPU coprocessor\u003c\/strong\u003e to your system, enabling high-speed machine learning inferencing on a wide range of systems, simply by connecting it to a USB port!\u003c\/p\u003e\n\u003cp class=\" small-headline mk-heading\" id=\"performs-high-speed-ml-inferencing\"\u003e\u003cstrong\u003ePerforms High-speed Machine Learning Inferencing\u003c\/strong\u003e\u003ci class=\"material-icons icon-link\"\u003e\u003c\/i\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eThe on-board Edge TPU coprocessor is capable of performing 4 trillion operations (tera-operations) per second (TOPS), using 0.5 watts for each TOPS (2 TOPS per watt). For example, it can execute state-of-the-art mobile vision models such as MobileNet v2 at almost 400 FPS, in a power-efficient manner.\u003cspan\u003e See below section for performance benchmarks.\u003c\/span\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e\u003cspan\u003e\u003c\/span\u003e\u003cstrong\u003eSupports all major platforms\u003c\/strong\u003e\u003ci class=\"material-icons icon-link\"\u003e\u003c\/i\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eConnects via USB to any system running Debian Linux (including Raspberry Pi), macOS, or Windows 10.\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e\u003cstrong\u003eSupports TensorFlow Lite\u003c\/strong\u003e\u003ci class=\"material-icons icon-link\"\u003e\u003c\/i\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eNo need to build models from the ground up.\u003cspan\u003e \u003c\/span\u003eTensorFlow Lite\u003cspan\u003e \u003c\/span\u003emodels can be compiled to run on the Edge TPU.\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e\u003cstrong\u003eSupports AutoML Vision Edge\u003c\/strong\u003e\u003ci class=\"material-icons icon-link\"\u003e\u003c\/i\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eEasily build and deploy fast, high-accuracy custom image classification models to your device with\u003cspan\u003e \u003c\/span\u003eAutoML Vision Edge.\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e\u003cstrong\u003eTech specs\u003c\/strong\u003e\u003ci class=\"material-icons icon-link\"\u003e\u003c\/i\u003e\u003c\/p\u003e\n\u003ctable class=\"table-group-table\" style=\"height: 72px;\" border=\"1\"\u003e\n\u003ctbody\u003e\n\u003ctr style=\"height: 36px;\"\u003e\n\u003ctd class=\"bold col-mobile-6 col-tablet-8 col-small-desktop-7 col-desktop-8\" style=\"height: 36px; width: 99px;\"\u003eML accelerator\u003c\/td\u003e\n\u003ctd style=\"height: 36px; width: 206px;\"\u003eGoogle Edge TPU coprocessor:\u003cbr\u003e4 TOPS (int8); 2 TOPS per watt\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"height: 18px;\"\u003e\n\u003ctd class=\"bold col-mobile-6 col-tablet-8 col-small-desktop-7 col-desktop-8\" style=\"height: 18px; width: 99px;\"\u003eConnector\u003c\/td\u003e\n\u003ctd style=\"height: 18px; width: 206px;\"\u003eUSB 3.0 Type-C* (data\/power)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"height: 18px;\"\u003e\n\u003ctd class=\"bold col-mobile-6 col-tablet-8 col-small-desktop-7 col-desktop-8\" style=\"height: 18px; width: 99px;\"\u003eDimensions\u003c\/td\u003e\n\u003ctd style=\"height: 18px; width: 206px;\"\u003e65 mm x 30 mm\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e* Compatible with USB 2.0 but inferencing speed is much slower.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eDatasheet \u0026amp; Resources\u003c\/strong\u003e\u003c\/p\u003e\n\u003cul\u003e\n\u003cli\u003eDatasheet\u003c\/li\u003e\n\u003cli\u003eGetting started with the USB Accelerator\u003c\/li\u003e\n\u003cli\u003eUSB Accelerator datasheet\u003c\/li\u003e\n\u003cli\u003eCoral Edge TPU Frequently asked questions (FAQ)\u003c\/li\u003e\n\u003cli\u003e\nEdge TPU Python API overview\n\u003cul\u003e\n\u003cli\u003eNote: this is NOT the TensorFlow Lite API, but an alternative API intended for users who have not used TensorFlow before and simply want to start with image classification and object detection\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/li\u003e\n\u003cli\u003e3D CAD model (.STEP file)\u003c\/li\u003e\n\u003cli\u003eEdge TPU inferencing overview\u003c\/li\u003e\n\u003cli\u003eTensorFlow models on the Edge TPU\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp\u003e\u003cstrong\u003ePerformance Benchmarks\u003c\/strong\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eAn individual Edge TPU is capable of performing 4 trillion operations (tera-operations) per second (TOPS), using 0.5 watts for each TOPS (2 TOPS per watt). How that translates to performance for your application depends on a variety of factors. Every neural network model has different demands, and \u003cstrong\u003eif you're using the USB Accelerator device, total performance also varies based on the host CPU, USB speed, and other system resources.\u003c\/strong\u003e\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eWith that said, the table below compares the time spent to perform a single inference with several popular models on the Edge TPU. For the sake of comparison, all models running on both CPU and Edge TPU are the TensorFlow Lite versions.\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003eThis represents a small selection of model architectures that are compatible with the Edge TPU:\u003c\/p\u003e\n\u003cp class=\" body-copy mk-paragraph\"\u003e\u003cb\u003eNote:\u003c\/b\u003e\u003cspan\u003e These figures measure the time required to execute the model only. It does not include the time to process input data (such as down-scaling images to fit the input tensor), which can vary between systems and applications. These tests are also performed using C++ benchmark tests, whereas our public \u003c\/span\u003ePython benchmark scripts\u003cspan\u003e may be slower due to overhead from Python.\u003c\/span\u003e\u003c\/p\u003e\n\u003ctable border=\"1\"\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003e\u003cstrong\u003eModel architecture\u003c\/strong\u003e\u003c\/th\u003e\n\u003cth class=\"blue\"\u003e\u003cstrong\u003eDesktop CPU\u003cspan\u003e \u003c\/span\u003e\u003csup\u003e1\u003c\/sup\u003e\u003c\/strong\u003e\u003c\/th\u003e\n\u003cth class=\"blue\"\u003e\n\u003cstrong\u003eDesktop CPU\u003cspan\u003e \u003c\/span\u003e\u003csup\u003e1\u003c\/sup\u003e\u003cbr\u003e+ USB Accelerator (USB 3.0)\u003c\/strong\u003e\u003cbr\u003e\u003cspan class=\"notice\"\u003ewith Edge TPU\u003c\/span\u003e\n\u003c\/th\u003e\n\u003cth class=\"green\"\u003e\u003cstrong\u003eEmbedded CPU\u003cspan\u003e \u003c\/span\u003e\u003csup\u003e2\u003c\/sup\u003e\u003c\/strong\u003e\u003c\/th\u003e\n\u003cth class=\"green\"\u003e\n\u003cstrong\u003eDev Board\u003cspan\u003e \u003c\/span\u003e\u003csup\u003e3\u003c\/sup\u003e\u003c\/strong\u003e\u003cbr\u003e\u003cspan class=\"notice\"\u003ewith Edge TPU\u003c\/span\u003e\n\u003c\/th\u003e\n\u003c\/tr\u003e\n\u003c\/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd\u003eUnet Mv2\u003cbr\u003e(128x128)\u003c\/td\u003e\n\u003ctd\u003e27.7\u003c\/td\u003e\n\u003ctd\u003e3.3\u003c\/td\u003e\n\u003ctd\u003e190.7\u003c\/td\u003e\n\u003ctd\u003e5.7\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eDeepLab V3\u003cbr\u003e(513x513)\u003c\/td\u003e\n\u003ctd\u003e394\u003c\/td\u003e\n\u003ctd\u003e52\u003c\/td\u003e\n\u003ctd\u003e1139\u003c\/td\u003e\n\u003ctd\u003e241\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eDenseNet\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e380\u003c\/td\u003e\n\u003ctd\u003e20\u003c\/td\u003e\n\u003ctd\u003e1032\u003c\/td\u003e\n\u003ctd\u003e25\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eInception v1\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e90\u003c\/td\u003e\n\u003ctd\u003e3.4\u003c\/td\u003e\n\u003ctd\u003e392\u003c\/td\u003e\n\u003ctd\u003e4.1\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eInception v4\u003cbr\u003e(299x299)\u003c\/td\u003e\n\u003ctd\u003e700\u003c\/td\u003e\n\u003ctd\u003e85\u003c\/td\u003e\n\u003ctd\u003e3157\u003c\/td\u003e\n\u003ctd\u003e102\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eInception-ResNet V2\u003cbr\u003e(299x299)\u003c\/td\u003e\n\u003ctd\u003e753\u003c\/td\u003e\n\u003ctd\u003e57\u003c\/td\u003e\n\u003ctd\u003e2852\u003c\/td\u003e\n\u003ctd\u003e69\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eMobileNet v1\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e53\u003c\/td\u003e\n\u003ctd\u003e2.4\u003c\/td\u003e\n\u003ctd\u003e164\u003c\/td\u003e\n\u003ctd\u003e2.4\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eMobileNet v2\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e51\u003c\/td\u003e\n\u003ctd\u003e2.6\u003c\/td\u003e\n\u003ctd\u003e122\u003c\/td\u003e\n\u003ctd\u003e2.6\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eMobileNet v1 SSD\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e109\u003c\/td\u003e\n\u003ctd\u003e6.5\u003c\/td\u003e\n\u003ctd\u003e353\u003c\/td\u003e\n\u003ctd\u003e11\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eMobileNet v2 SSD\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e106\u003c\/td\u003e\n\u003ctd\u003e7.2\u003c\/td\u003e\n\u003ctd\u003e282\u003c\/td\u003e\n\u003ctd\u003e14\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eResNet-50 V1\u003cbr\u003e(299x299)\u003c\/td\u003e\n\u003ctd\u003e484\u003c\/td\u003e\n\u003ctd\u003e49\u003c\/td\u003e\n\u003ctd\u003e1763\u003c\/td\u003e\n\u003ctd\u003e56\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eResNet-50 V2\u003cbr\u003e(299x299)\u003c\/td\u003e\n\u003ctd\u003e557\u003c\/td\u003e\n\u003ctd\u003e50\u003c\/td\u003e\n\u003ctd\u003e1875\u003c\/td\u003e\n\u003ctd\u003e59\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eResNet-152 V2\u003cbr\u003e(299x299)\u003c\/td\u003e\n\u003ctd\u003e1823\u003c\/td\u003e\n\u003ctd\u003e128\u003c\/td\u003e\n\u003ctd\u003e5499\u003c\/td\u003e\n\u003ctd\u003e151\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eSqueezeNet\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e55\u003c\/td\u003e\n\u003ctd\u003e2.1\u003c\/td\u003e\n\u003ctd\u003e232\u003c\/td\u003e\n\u003ctd\u003e2\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eVGG16\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e867\u003c\/td\u003e\n\u003ctd\u003e296\u003c\/td\u003e\n\u003ctd\u003e4595\u003c\/td\u003e\n\u003ctd\u003e343\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eVGG19\u003cbr\u003e(224x224)\u003c\/td\u003e\n\u003ctd\u003e1060\u003c\/td\u003e\n\u003ctd\u003e308\u003c\/td\u003e\n\u003ctd\u003e5538\u003c\/td\u003e\n\u003ctd\u003e357\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eEfficientNet-EdgeTpu-S*\u003c\/td\u003e\n\u003ctd\u003e5431\u003c\/td\u003e\n\u003ctd\u003e5.1\u003c\/td\u003e\n\u003ctd\u003e705\u003c\/td\u003e\n\u003ctd\u003e5.5\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eEfficientNet-EdgeTpu-M*\u003c\/td\u003e\n\u003ctd\u003e8469\u003c\/td\u003e\n\u003ctd\u003e8.7\u003c\/td\u003e\n\u003ctd\u003e1081\u003c\/td\u003e\n\u003ctd\u003e10.6\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003eEfficientNet-EdgeTpu-L*\u003c\/td\u003e\n\u003ctd\u003e22258\u003c\/td\u003e\n\u003ctd\u003e25.3\u003c\/td\u003e\n\u003ctd\u003e2717\u003c\/td\u003e\n\u003ctd\u003e30.5\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003cp class=\" body-copy mk-paragraph small-headline\"\u003e\u003csup\u003e1\u003c\/sup\u003e\u003cspan\u003e \u003c\/span\u003e\u003ci\u003eDesktop CPU: Single 64-bit Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz\u003c\/i\u003e\u003cbr\u003e\u003csup\u003e2\u003c\/sup\u003e\u003cspan\u003e \u003c\/span\u003e\u003ci\u003eEmbedded CPU: Quad-core Cortex-A53 @ 1.5GHz\u003c\/i\u003e\u003cbr\u003e\u003csup\u003e3\u003c\/sup\u003e\u003cspan\u003e \u003c\/span\u003e\u003ci\u003eDev Board: Quad-core Cortex-A53 @ 1.5GHz + Edge TPU\u003c\/i\u003e\u003c\/p\u003e\n\u003cp\u003e \u003c\/p\u003e","brand":"Google","offers":[{"title":"Default Title","offer_id":52217475367200,"sku":"ysp1787258097253","price":68.75,"currency_code":"USD","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0946\/3231\/3120\/files\/coral-usb-accelerator_1.jpg?v=1787258101","url":"https:\/\/www.elmwoodbasket.shop\/products\/coral-usb-accelerator","provider":"My Store","version":"1.0","type":"link"}