--- title: Build a Real-Time Video Analytics Pipeline with YOLO26... description: This tutorial builds a complete real-time video analytics pipeline combining YOLO26 object detection with ByteTrack multi-object tracking for people and... canonical: https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack | daily.dev og:description: This tutorial builds a complete real-time video analytics pipeline combining YOLO26 object detection with ByteTrack multi-object tracking for people and... og:url: https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a og:image: https://api.daily.dev/og/posts/W83NrKa6a.png og:image:alt: Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack **[PyImageSearch](https://daily.dev/sources/pyimagesearch)** · 30 min read · 0 upvotes · 0 comments ## Summary This tutorial builds a complete real-time video analytics pipeline combining YOLO26 object detection with ByteTrack multi-object tracking for people and vehicle counting on macOS. The pipeline flow is Video -> YOLO26 -> ByteTrack -> Analytics -> Output, using Python packages ultralytics, supervision, and opencv-python. A line-crossing analytics engine tracks stable object IDs to count crossings, with presets for people counting (one line at 55% frame height, person class) and traffic counting (two lines at 42% and 68%, vehicle classes). On Apple Silicon, MPS delivers about 3.7 times the throughput of CPU execution under the tested settings. The code also supports zone occupancy counting, multi-line configs, and frame resizing or skipping for performance tuning. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much faster is MPS than CPU for running YOLO26 video inference on Apple Silicon? MPS delivers approximately 3.7 times the throughput of CPU execution under tested settings. This makes real-time performance achievable on Apple Silicon using Metal Performance Shaders, while CPU-only execution may struggle to keep up with higher-resolution video streams. The device toggle between 'cpu' and 'mps' is a key deployment lever for balancing speed and stability. _Comparing CPU and MPS throughput for vision models is easier with benchmarks surfaced on daily.dev._ ### Why use ByteTrack instead of a simple centroid tracker for counting objects in video? ByteTrack handles occlusion and brief disappearances better than naive centroid matching, uses detection confidence during association, and maintains a lost-track buffer that helps in busy scenes. This makes it a better fit for real-world people counting and vehicle counting pipelines where track IDs need to stay stable under noise and partial visibility, compared to detection-only or centroid-based approaches. _Developers choosing a tracking algorithm for a counting pipeline can follow these comparisons on daily.dev._ ### How do you avoid double-counting the same object crossing a counting line twice in a video analytics pipeline? Each tracked object is counted only once per line by adding its track ID to a counted_track_ids set after the first crossing, so the same object is not counted repeatedly. This count-once logic works well for entry and exit use cases but means a return crossing of the same track across the same line is not counted; handling repeated crossings needs a more advanced per-track state machine. _Engineers implementing line-crossing counters can track implementation details like this on daily.dev._ ## Similar posts on daily.dev - [How to Build a Real-Time Object Detection and Tracking Pipeline with ROS 2 and YOLOv11](https://daily.dev/posts/how-to-build-a-real-time-object-detection-and-tracking-pipeline-with-ros-2-and-yolov11-uipmgqo4d) · freeCodeCamp · 1 upvotes · 0 comments - [Ultralytics Optimizes YOLO26 for Intel OpenVINO](https://daily.dev/posts/ultralytics-optimizes-yolo26-for-intel-openvino-ljsupzlmx) · Embedded.com · 1 upvotes · 0 comments - [Train YOLO26 for Retail Object Detection on DigitalOcean GPUs](https://daily.dev/posts/train-yolo26-for-retail-object-detection-on-digitalocean-gpus-8e9osvjze) · DigitalOcean Community · 0 upvotes · 0 comments - [Researchers Solved a Decade-old Problem in Object Detection](https://daily.dev/posts/researchers-solved-a-decade-old-problem-in-object-detection-nr8nvuooe) · Daily Dose of Data Science \| Avi Chawla \| Substack · 2 upvotes · 0 comments - [YOLO26 Open-Vocabulary Object Detection with YOLOE-26](https://daily.dev/posts/yolo26-open-vocabulary-object-detection-with-yoloe-26-rpj3kfesx) · PyImageSearch · 1 upvotes · 0 comments --- Tags: [#computer-vision](https://daily.dev/tags/computer-vision), [#opencv](https://daily.dev/tags/opencv) [View this post on daily.dev](https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack","url":"https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a"},"datePublished":"2026-11-02T13:49:25.135Z","dateModified":"2026-11-02T13:49:55.123Z","description":"This tutorial builds a complete real-time video analytics pipeline combining YOLO26 object detection with ByteTrack multi-object tracking for people and...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5dca9b85a0559602080bdd7eb0a09376?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5dca9b85a0559602080bdd7eb0a09376?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"PyImageSearch","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"PyImageSearch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/33ddc48b99ee47eb9283ef1fdcff3095","url":"https://daily.dev/sources/pyimagesearch"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"computer-vision,opencv","timeRequired":"PT30M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"PyImageSearch","item":"https://daily.dev/sources/pyimagesearch"},{"@type":"ListItem","position":3,"name":"Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/build-a-real-time-video-analytics-pipeline-with-yolo26-and-bytetrack-w83nrka6a#faq","mainEntity":[{"@type":"Question","name":"How much faster is MPS than CPU for running YOLO26 video inference on Apple Silicon?","acceptedAnswer":{"@type":"Answer","text":"MPS delivers approximately 3.7 times the throughput of CPU execution under tested settings. This makes real-time performance achievable on Apple Silicon using Metal Performance Shaders, while CPU-only execution may struggle to keep up with higher-resolution video streams. The device toggle between 'cpu' and 'mps' is a key deployment lever for balancing speed and stability. Comparing CPU and MPS throughput for vision models is easier with benchmarks surfaced on daily.dev."}},{"@type":"Question","name":"Why use ByteTrack instead of a simple centroid tracker for counting objects in video?","acceptedAnswer":{"@type":"Answer","text":"ByteTrack handles occlusion and brief disappearances better than naive centroid matching, uses detection confidence during association, and maintains a lost-track buffer that helps in busy scenes. This makes it a better fit for real-world people counting and vehicle counting pipelines where track IDs need to stay stable under noise and partial visibility, compared to detection-only or centroid-based approaches. Developers choosing a tracking algorithm for a counting pipeline can follow these comparisons on daily.dev."}},{"@type":"Question","name":"How do you avoid double-counting the same object crossing a counting line twice in a video analytics pipeline?","acceptedAnswer":{"@type":"Answer","text":"Each tracked object is counted only once per line by adding its track ID to a counted_track_ids set after the first crossing, so the same object is not counted repeatedly. This count-once logic works well for entry and exit use cases but means a return crossing of the same track across the same line is not counted; handling repeated crossings needs a more advanced per-track state machine. Engineers implementing line-crossing counters can track implementation details like this on daily.dev."}}]} ```