--- format: "aidr-story-markdown/v1" id: "932dee212b4ccfaabd460e86091915c5e3a56884b7fb05b47f7234ecc8025c8e" canonical_url: "https://aidr.today/932dee21?lang=en" title: "Perceptron Launches EgoHome Dataset for Fine-Grained Robot Vision Benchmarking" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-11-10T18:53:48.000Z" category: "Data" topics: ["open-source","data-engineering","openai","llm"] source_urls: ["https://marketbrief.now/ai/perceptron-launches-egohome-dataset-for-fine-grained-robot-vision-benchm-841f87cc","https://huggingnews.com/ai/perceptron-launches-egohome-dataset-for-fine-grained-robot-vision-benchm-841f87cc","https://huggingnews.com/ai/update-gpt-6-astra-leads-performance-on-perceptron-ais-new-robot-vision-8b0fe16a","https://marketbrief.now/ai/update-gpt-6-astra-leads-performance-on-perceptron-ais-new-robot-vision-8b0fe16a"] summary: "Perceptron has open-sourced the EgoHome benchmark. This benchmark includes 2 hours of human-annotated video featuring more than 750 video segments that last less than 500ms. The dataset provides four specific annotation types for every clip." --- # Perceptron Launches EgoHome Dataset for Fine\-Grained Robot Vision Benchmarking > [Open the canonical story]() **Published:** 2026-11-10T18:53:48.000Z **Category:** Data **Topics:** open\-source, data\-engineering, openai, llm ## Summary Perceptron has open\-sourced the EgoHome benchmark\. This benchmark includes 2 hours of human\-annotated video featuring more than 750 video segments that last less than 500ms\. The dataset provides four specific annotation types for every clip\. ## Sources - [Story source]() - [Story source]() - [Story source]() - [Story source]()