EgoHome: open egocentric benchmark with 2 hours of annotated household-chore video
AkshatS07 · x · 2026-11-11
Perceptron released EgoHome, an open benchmark for dense, sub-second egocentric understanding:
- 2 hours of human-annotated household-chore video, with 4 annotation types on every clip: hand boxes, subtasks, hand actions, and hand captions.
- Over 750 segments under 500ms; data, metrics, and judge prompts are all open.
- The hardest part was training annotators and building scalable, cost-effective annotation UX; the author personally spent 40 hours annotating.
- An object-tracking addition to the benchmark is planned for release in a few weeks.
Data is on Hugging Face; blog and repo links in the original post.
More from Research
- TRIAD benchmark: no multimodal retrieval system leads across its six subsets — _reachsumit · 2026-11-11
- Vision tokens far denser than text: paper traces VLM modality gap to early-layer processing — _reachsumit · 2026-11-11
- ByteDance's SepLRM splits recommendation models around a shared KV cache — _reachsumit · 2026-11-11
- Yuanbao-Reranker: 0.6B/4B/8B Qwen3-based rerankers trained on synthetic preferences, open-sourced — _reachsumit · 2026-11-11
- Shopify's RoutingTrie lifts generative retrieval Hit@100 from 67.5% to 96.1% with 1.1ms overhead — _reachsumit · 2026-11-11
- Hybrid product search straight from the data lake: Pinot + BM25 + RRF hits 6.5-9ms p50 on 277M products — _reachsumit · 2026-11-11