https://www.pipeless.ai/docs/_next/static/chunks/pages/v1/formats-98533935dff1d69d.js

1(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[460],{7522:function(e,t,o){(window.__NEXT_P=window.__NEXT_P||[]).push(["/v1/formats",function(){return o(5205)}])},9369:function(e,t,o){"use strict";var n=o(5893);o(7294);var i=o(5675),s=o.n(i),r=o(3750),a=o(1163),d=o(2681);let l={logo:(0,n.jsxs)("span",{style:{display:"flex",alignItems:"center"},children:[(0,n.jsx)(s(),{alt:"logo",width:30,height:30,src:"/docs/logo-400x400-rounded.png"}),"\xa0",(0,n.jsx)("b",{children:"Pipeless"})]}),logoLink:"https://pipeless.ai",primaryHue:297,project:{link:"https://github.com/pipeless-ai/pipeless"},chat:{link:"https://discord.gg/K2qxQ8uedG",icon:(0,n.jsx)(r.KpZ,{size:25})},docsRepositoryBase:"https://github.com/pipeless-ai/docs/tree/main",footer:{text:(0,n.jsx)("p",{children:"Copyright \xa9 2023 Pipeless, Inc."})},search:{loading:"Loading..."},head:()=>{let{asPath:e,defaultLocale:t,locale:o}=(0,a.useRouter)(),{frontMatter:i}=(0,d.ZR)(),s="https://www.pipeless.ai"+(t===o?e:"/".concat(o).concat(e));return(0,n.jsxs)(n.Fragment,{children:[(0,n.jsx)("meta",{property:"og:url",content:s}),(0,n.jsx)("meta",{property:"og:title",content:i.title||"Pipeless Documentation"}),(0,n.jsx)("meta",{property:"og:description",content:i.description||"Documentation about Pipeless, the computer vision framework"}),(0,n.jsx)("meta",{property:"og:image",content:"https://www.pipeless.ai/docs/logo-400x400-rounded.png"})]})},useNextSeoProps:()=>({titleTemplate:"%s – Pipeless"}),banner:{dismissible:!1,key:"agents",text:(0,n.jsx)("a",{target:"blank",href:"https://agents.pipeless.ai",children:(0,n.jsxs)("p",{children:["\uD83D\uDE80 \xa0\xa0",(0,n.jsx)("b",{children:"Pipeless Agents is live!"}),"\xa0Create vision powered apps and automations effortlessly within seconds.\xa0",(0,n.jsx)("span",{style:{textDecoration:"underline"},children:"Click here to get started!"})]})})},sidebar:{defaultMenuCollapseLevel:1}};t.Z=l},5205:function(e,t,o){"use strict";o.r(t),o.d(t,{__toc:function(){return l}});var n=o(5893),i=o(2673),s=o(2681),r=o(9369);o(9966);var a=o(2643),d=o(9013);let l=[{depth:2,value:"Computer vision / video processing",id:"computer-vision--video-processing"},{depth:3,value:"Input",id:"input"},{depth:4,value:"Select input camera",id:"select-input-camera"},{depth:3,value:"Output",id:"output"},{depth:2,value:"Audio recognition / audio processing",id:"audio-recognition--audio-processing"},{depth:2,value:"Data and Events",id:"data-and-events"}];function c(e){let t=Object.assign({h1:"h1",p:"p",strong:"strong",a:"a",h2:"h2",h3:"h3",code:"code",h4:"h4",table:"table",thead:"thead",tr:"tr",th:"th",tbody:"tbody",td:"td"},(0,a.a)(),e.components);return(0,n.jsxs)(n.Fragment,{children:[(0,n.jsx)(t.h1,{children:"Supported Stream Formats and Protocols"}),"\n",(0,n.jsxs)(t.p,{children:["Even though Pipeless is defined as a computer vision framework, it is designed to allow you to analyze any kind of stream,\nincluding video, audio, subtitles, etc. These streams can be ",(0,n.jsx)(t.strong,{children:"analyzed and modified in real-time"}),".\nWhen analyzing a stream you can ",(0,n.jsx)(t.strong,{children:"export data and events"})," and when modifying you can generate a new stream with some changes and\nthe same encoding and format as the original one."]}),"\n","\n",(0,n.jsx)(d.UW,{children:(0,n.jsxs)(t.p,{children:["If you need a format that is not supported please open a ",(0,n.jsx)(t.a,{href:"https://github.com/pipeless-ai/pipeless/issues/new?title=Feature%20Request%20%E2%80%9CYour%20Feature&labels=feature",children:"feature request"})," or if you implemented a new format you can contribute by creating a pull request on GitHub."]})}),"\n",(0,n.jsx)(t.h2,{id:"computer-vision--video-processing",children:"Computer vision / video processing"}),"\n",(0,n.jsx)(t.h3,{id:"input",children:"Input"}),"\n",(0,n.jsxs)(t.p,{children:["Pipeless supports almost ",(0,n.jsx)(t.strong,{children:"any protocol and format"})," (with several codecs) as input."]}),"\n",(0,n.jsxs)(t.p,{children:["Supported input protocols: ",(0,n.jsx)(t.code,{children:"v4l2"})," (for device webcam), ",(0,n.jsx)(t.code,{children:"file"}),", ",(0,n.jsx)(t.code,{children:"http(s)"}),", ",(0,n.jsx)(t.code,{children:"rtmp"}),", ",(0,n.jsx)(t.code,{children:"rtsp"}),", ",(0,n.jsx)(t.code,{children:"rtp"}),", ",(0,n.jsx)(t.code,{children:"tcp"}),", ",(0,n.jsx)(t.code,{children:"udp"}),", ",(0,n.jsx)(t.code,{children:"ftp"}),", ...\nSupported input formats: ",(0,n.jsx)(t.code,{children:"mp4"}),", ",(0,n.jsx)(t.code,{children:"webm"}),", ",(0,n.jsx)(t.code,{children:"mkv"}),", ... (several codecs supported for all of them)"]}),"\n",(0,n.jsx)(t.h4,{id:"select-input-camera",children:"Select input camera"}),"\n",(0,n.jsxs)(t.p,{children:["You can use ",(0,n.jsx)(t.code,{children:"v4l2:/dev/videoX"})," where ",(0,n.jsx)(t.code,{children:"X"})," is the index of the device as input URI to select the input device when you have multiple cameras. For example, ",(0,n.jsx)(t.code,{children:"v4l2:/dev/video0"}),". If you provide ",(0,n.jsx)(t.code,{children:"v4l2"})," without specifying the device it will fallback to ",(0,n.jsx)(t.code,{children:"/dev/video0"}),"."]}),"\n",(0,n.jsx)(t.h3,{id:"output",children:"Output"}),"\n",(0,n.jsx)(t.p,{children:"When the output video is enabled, the following table describes the supported combinations of protocols and formats. New output protocols and formats are added constantly."}),"\n",(0,n.jsxs)(t.table,{children:[(0,n.jsx)(t.thead,{children:(0,n.jsxs)(t.tr,{children:[(0,n.jsx)(t.th,{children:"Output Protocol"}),(0,n.jsx)(t.th,{children:"Output Format"})]})}),(0,n.jsxs)(t.tbody,{children:[(0,n.jsxs)(t.tr,{children:[(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"screen"})}),(0,n.jsxs)(t.td,{children:[(0,n.jsx)(t.code,{children:"raw"})," (Directly shown on the device screen)"]})]}),(0,n.jsxs)(t.tr,{children:[(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"file"})}),(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"mp4"})})]}),(0,n.jsxs)(t.tr,{children:[(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"rtmp"})}),(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"flv"})})]}),(0,n.jsxs)(t.tr,{children:[(0,n.jsx)(t.td,{children:(0,n.jsx)(t.code,{children:"rtsp"})}),(0,n.jsxs)(t.td,{children:["Same as input. ",(0,n.jsx)(t.code,{children:"h.264"})," encoded"]})]})]})]}),"\n",(0,n.jsx)(d.UW,{children:(0,n.jsxs)(t.p,{children:["NOTE: When using ",(0,n.jsx)(t.code,{children:"v4l2"})," as input URI to read from the device webcam, the output URI (if the output is enabled) ",(0,n.jsx)(t.strong,{children:"must"})," be ",(0,n.jsx)(t.code,{children:"screen"}),". ",(0,n.jsx)(t.code,{children:"v4l2"})," as input is supported in core versions >= ",(0,n.jsx)(t.code,{children:"0.1.3-alpha"})]})}),"\n",(0,n.jsx)(t.h2,{id:"audio-recognition--audio-processing",children:"Audio recognition / audio processing"}),"\n",(0,n.jsx)(t.p,{children:"Audio processing is being implemented, however, not yet available."}),"\n",(0,n.jsx)(t.h2,{id:"data-and-events",children:"Data and Events"}),"\n",(0,n.jsxs)(t.p,{children:["You can easily export data or handle events in the ",(0,n.jsx)(t.code,{children:"post-process"})," hooks of your stages as you would usually do in any other application."]})]})}let p={MDXContent:function(){let e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{},{wrapper:t}=Object.assign({},(0,a.a)(),e.components);return t?(0,n.jsx)(t,{...e,children:(0,n.jsx)(c,{...e})}):c(e)},pageOpts:{filePath:"pages/v1/formats.mdx",route:"/v1/formats",frontMatter:{title:"Supported stream formats and protocols",description:"Learn the supported formats and protocols that can be used to provide video to your Pipeless application"},timestamp:1705418408e3,pageMap:[{kind:"Meta",data:{index:"Introduction",contributing:"Contributing","-- Versions":{type:"separator",title:"Versions"},v1:"v1.x.x",v0:"v0.x.x",about:{title:"Home",type:"page",href:"https://pipeless.ai"},contact:{title:"Contact ↗",type:"page",href:"https://lfkz9noht13.typeform.com/to/aeOqKGPt",newWindow:!0}}},{kind:"MdxPage",name:"contributing",route:"/contributing",frontMatter:{title:"Contributing Guidelines",description:"Contribute to the Pipeless open-source computer vision framework in GitHub"}},{kind:"MdxPage",name:"index",route:"/",frontMatter:{title:"Introduction",description:"Introduction to Pipeless, the open-source and production ready computer vision framework. Enjoy serverless development experience for embedded computer vision."}},{kind:"MdxPage",name:"v0",route:"/v0",frontMatter:{title:"Pipeless v0.x",description:"Documentation for Pipeless ver

1sions 0.x. The open-source computer vision framework"}},{kind:"Folder",name:"v1",route:"/v1",children:[{kind:"Meta",data:{"getting-started":"\uD83D\uDE80 Getting Started","rest-api":"Pipeless REST API",formats:"Stream Formats and Protocols",inference:"Inference Runtimes",kvstore:"Key Value Store",container:"\uD83D\uDC0B Container images",deployment:"Deploying your application",models:"Ready to Use CV Models",examples:"\uD83C\uDF1F Examples",benchmark:"Benchmark",changes:"Version Notable Changes",troubleshooting:"Troubleshooting","restart-policy":"Processing Restart Policy"}},{kind:"MdxPage",name:"benchmark",route:"/v1/benchmark",frontMatter:{title:"Benchmark",description:"A benchmark obtained by running the Pipeless computer vision framework on different devices to measure the overhead"}},{kind:"MdxPage",name:"changes",route:"/v1/changes",frontMatter:{title:"Notable version changes",description:"A list of important changes between Pipeless versions"}},{kind:"MdxPage",name:"container",route:"/v1/container",frontMatter:{title:"Container Images",description:"Learn to run and deploy Pipeless using container images. There are images available for CPU, and GPU (CUDA and TensorRT). The container images can also be used to easily deploy Pipeless to the cloud."}},{kind:"MdxPage",name:"deployment",route:"/v1/deployment",frontMatter:{title:"Deploying your pipeless application",description:"Learn how to deploy a Pipeless application to either an edge device or to the cloud. Pipeless ca run in distributed cloud servers as well as edge and IoT devices."}},{kind:"Folder",name:"examples",route:"/v1/examples",children:[{kind:"Meta",data:{w

1attermark:"Adding wattermarks",cats:"Detecting Cats on a Video","tf-pose":"Pose detection with TensorFlow",yolov8:"Object detection - YOLOv8 Python lib","onnx-candy":"ONXX Runtime - Candy filter","onnx-yolo":"ONNX Runtime - Object detection with YOLO","gaze-piano":"Play a piano with your eyes by looking to the notes"}},{kind:"MdxPage",name:"cats",route:"/v1/examples/cats",frontMatter:{title:"Cats recognition example",description:"A Pipeless example to recognize and detect cats in a video stream"}},{kind:"MdxPage",name:"gaze-piano",route:"/v1/examples/gaze-piano",frontMatter:{title:"Play a piano with your eyes - Gaze estimation",description:"A Pipeless example that allows you play a virtual piano with your eyes by looking to the note you want to play"}},{kind:"MdxPage",name:"onnx-candy",route:"/v1/examples/onnx-candy",frontMatter:{title:"Candy filter with ONNX runtime",description:"A Pipeless example using the ONNX Runtime to apply a candy filter to a video stream"}},{kind:"MdxPage",name:"onnx-yolo",route:"/v1/examples/onnx-yolo",frontMatter:{title:"Object detection with ONNX Runtime",description:"A Pipeless example that runs inference using the ONNX Runtime and YOLO to detect objects in a video stream"}},{kind:"MdxPage",name:"tf-pose",route:"/v1/examples/tf-pose",frontMatter:{title:"Pose estimation/detection with TensorFlow",description:"A Pipeless example that detects the pose with 17 keypoints in a video stream"}},{kind:"MdxPage",name:"wattermark",route:"/v1/examples/wattermark",frontMatter:{title:"Adding wattermarks to a video",description:"The most basic Pipeless example adding a wattermark to a video"}},{kind:"MdxPage",name:"yolov8",route:"/v1/examples/yolov8",frontMatter:{title:"Object detection with YOLOv8 and Ultralytics SDK",description:"A Pipeless example to detect objects in a video stream using YOLOv8 via the Ultralytics Python SDK"}}]},{kind:"MdxPage",name:"examples",route:"/v1/examples",frontMatter:{title:"Examples",description:"A list of fully functional examples to play around with the Pipeless computer vision framework."}},{kind:"MdxPage",name:"formats",route:"/v1/formats",frontMatter:{title:"Supported stream formats and protocols",description:"Learn the supported formats and protocols that can be used to provide video to your Pipeless application"}},{kind:"Folder",name:"getting-started",route:"/v1/getting-started",children:[{kind:"Meta",data:{installation:"\uD83D\uDEE0️ Installation","project-structure":"\uD83D\uDCC2 Project Structure"}},{kind:"MdxPage",name:"installation",route:"/v1/getting-started/installation",frontMatter:{title:"Installation guide",description:"Learn to install Pipeless, the computer vision framework"}},{kind:"MdxPage",name:"project-structure",route:"/v1/getting-started/project-structure",frontMatter:{title:"Project structure",description:"Learn how a Pipeless project is structured in the filesystem"}}]},{kind:"MdxPage",name:"getting-started",route:"/v1/getting-started",frontMatter:{title:"Getting started",description:"Get started with the Pipeless computer vision framework and enjoy the development experience of serverless frameworks for embedded computer vision."}},{kind:"MdxPage",name:"inference",route:"/v1/inference",frontMatter:{title:"Inference runtimes",description:"Learn to how to load a model into one of the embedded inference runtimes to perform your computer vision tasks in CPU, or GPUs. Use CUDA, TensorRT, OpenVINO, CoreML in any kind of GPU, including Nvidia GPUs."}},{kind:"MdxPage",name:"kvstore",route:"/v1/kvstore",frontMatter:{title:"Key-Value Store usage",description:"Learn to how use the included key value store to communicate data between the processing of different frames and the execution of different hooks"}},{kind:"Folder",name:"models",route:"/v1/models",children:[{kind:"Folder",name:"tensorflow",route:"/v1/models/tensorflow",children:[{kind:"MdxPage",name:"multi-pose-estimation",route:"/v1/models/tensorflow/multi-pose-estimation",frontMatter:{title:"Multi-pose estimation Tensorflow Model",description:"A ready to use TensorFlow model for multipose estimation in Python"}},{kind:"Meta",data:{"multi-pose-estimation":"Multi-pose estimation Tensorflow Model"}}]},{kind:"MdxPage",name:"tensorflow",route:"/v1/models/tensorflow"},{kind:"Meta",data:{tensorflow:"Tensorflow"}}]},{kind:"MdxPage",name:"models",route:"/v1/models",frontMatter:{title:"Ready to use computer vision models",description:"A list of ready to use computer vision models with Python APIs"}},{kind:"MdxPage",name:"rest-api",route:"/v1/rest-api",frontMatter:{title:"Pipeless REST API",description:"Documentation about the endpoints exposed by a running Pipeless instance"}},{kind:"MdxPage",name:"restart-policy",route:"/v1/restart-policy",frontMatter:{title:"Processing Restart Policy",description:"Documentation about the restart policy for stream processing"}},{kind:"MdxPage",name:"troubleshooting",route:"/v1/troubleshooting",frontMatter:{title:"Troubleshooting your Pipeless application",description:"General guidance and common issues when running your Pipeless application"}}]},{kind:"MdxPage",name:"v1",route:"/v1",frontMatter:{title:"Pipeless v1.x",description:"Documentation for Pipeless ver

1sions 1.x. The open-source computer vision framework"}}],flexsearch:{codeblocks:!0},title:"Supported stream formats and protocols",headings:l},pageNextRoute:"/v1/formats",nextraLayout:s.ZP,themeConfig:r.Z};t.default=(0,i.j)(p)}},function(e){e.O(0,[13,996,774,888,179],function(){return e(e.s=7522)}),_N_E=e.O()}]);