Perceptron launches Ego 1.0 API: egocentric video annotation at ~1/8 the cost of Astra
AkshatS07 · x · 2026-11-19
Perceptron has released Egocentric 1.0, a self-service API that turns first-person human and robot video into structured training data for robotics, priced at $29 per video-hour — less than half of GPT-6 Astra's $76 per video-hour for subtask timelines alone, with an overall claimed 8x cost advantage.
Key capabilities:
- Fine-grained subtask timelines matching Astra quality
- State-of-the-art 2D and 3D hand tracking, with a proprietary 3D hand model built on Mk1.5
- Independent continuous action timelines for each hand and robot gripper, including contact and visibility attributes
- Covers household chores, robot-arm sorting/arranging, and manufacturing tasks like circuit-board assembly and box packing; works on phone video without calibration
The team also says it will soon open-source EgoHome, a benchmark for fine-grained task annotation.
More from Research
- Hashimoto's Stanford CS312 quiz: what happens if you untie a looped Transformer mid-training? — tatsu_hashimoto · 2026-11-19
- Stanford CS312 experiment: looping transformers lose; mid-training untying recovers 61% of gap — tatsu_hashimoto · 2026-11-19
- Benchtop synthesis device releases have 3x'd; half of active manufacturers are in China — JacquesThibs · 2026-11-19
- Oxford paper: post-training makes LLM worldviews bleaker, models converge on 'spiritual bliss' — twiecki · 2026-11-19
- Researcher: Real scientific progress isn't rewarded, expect decades of incrementalism — gerardsans · 2026-11-19
- Grady Booch on LLM worldview paper: 'We made it exactly like we are, but worse' — Grady_Booch · 2026-11-19