---
title: README
emoji: 🧊
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
---

Macrodata Labs builds embodied data intelligence for robotics labs, researchers, and data vendors: transforming raw physical-world data into rich training datasets for robotics. Our optimized pipelines for automated QA, subtask annotation, egocentric hand tracking, and reward scoring extract the structured signals needed to train more capable robot policies.

🌐 [Website](https://macrodata.co?utm_source=huggingface&utm_medium=org_card&utm_campaign=org_readme&utm_content=useful_links) · 🐙 [GitHub](https://github.com/macrodata-labs/refiner) · 🐦 [Twitter](https://x.com/macrodata_labs) · 💬 [Discord](https://discord.gg/S8kZtmBR2x)

## Blog
- 🤖 [Segmenting Robot Video into Actionable Subtasks](https://macrodata.co/blog/annotating-robot-video-subtasks?utm_source=huggingface&utm_medium=org_card&utm_campaign=org_readme&utm_content=blog_list): how far off-the-shelf VLMs can push subtask annotation, and where they fail.
- 🖐️ [Turning Egocentric Video into Robot Actions](https://macrodata.co/blog/turning-egocentric-video-into-3d-hand-actions?utm_source=huggingface&utm_medium=org_card&utm_campaign=org_readme&utm_content=blog_list): what open-source hand tracking gets right on egocentric video, and where custom models are needed.

## Benchmarks
- 🥚 [WGO-Bench: What's Going On Benchmark](https://huggingface.co/datasets/macrodata/WGO-Bench): manually annotated robot and egocentric manipulation videos with timestamped subtask labels, for evaluating how well VLMs segment video into subtasks. Results are in our [write-up](https://macrodata.co/blog/annotating-robot-video-subtasks?utm_source=huggingface&utm_medium=org_card&utm_campaign=org_readme&utm_content=benchmarks).