# Perceptron — Introducing Perceptron Egocentric 1.0

- Company: Perceptron (perceptron.inc)
- Announced: 2026-11-04
- Category: api-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perceptron.inc/blog/introducing-perceptron-egocentric-1-0
- Record: https://forck.live/items/31482-introducing-perceptron-egocentric-1-0
- Subject: Perceptron Isaac / Perceptron Mk1
- Models affected: Perceptron Egocentric 1.0, GPT-6 Astra, FLORA, Pantheon Argus
- Pricing: $29 per video-hour; $9 per video-hour at 4 fps; estimated $15 at 8 fps; GPT-6 Astra ~$76 per video-hour

Perceptron is releasing Perceptron 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. The company says the API delivers frontier-quality subtask timelines, continuous activity labels for each hand, and 3D hand reconstruction through a single interface, with no manual onboarding. Perceptron says frontier models such as GPT-6 Astra annotate subtasks well but cost too much to run on large datasets, and that standard benchmarks miss segments shorter than a second. At its default 4 fps, Perceptron Egocentric scores 0.787 on EgoHome, the company's new benchmark of sub-second household steps, within 0.010 of GPT-6 Astra at 4 fps (0.797), for $9 per video-hour against roughly $76. At 8 fps it scores 0.798, level with GPT-6 Astra, at an estimated $15, and both settings outperform NVIDIA's FLORA (0.747). On the public WGO-Bench benchmark, Perceptron Egocentric achieves the highest segment F1 (0.583) and the second-highest semantic F1 (0.474). EgoHome consists of 137 manually verified clips covering roughly two hours of first-person household chores, 24 of which contain at least one step shorter than half a second; Perceptron says it will release the benchmark, a new permissible subsegment Jaccard metric and their code in the coming days.

## Evidence

Verbatim from https://www.perceptron.inc/blog/introducing-perceptron-egocentric-1-0:

> Today we're releasing Perceptron Egocentric 1.0 , a self-service API that turns first-person human and robot video into structured training data for robotics. It delivers frontier-quality subtask timelines, continuous activity labels for each hand, and state-of-the-art 3D hand reconstruction at $29 per video-hour, less than half of what GPT-6 Astra costs for subtask timelines alone (~$76 per video-hour), all accessible through a single API.
> 
> Dish washing, bare hands (phone video, no calibration): rinsing a frying pan while the other hand works the faucet with a sponge.
> 
> hold
> 
> frying pan handle
> 
> yes
> 
> visible
> 
> hold
> 
> blue sponge
> 
> yes
> 
> visible
> 
> Explore what Perceptron Egocentric produces on household, robot and manufacturing video: pick a clip, play it, click any step to jump to that moment, and switch between 2D keypoints and the 3D hand mesh. Robot clips are from the Galaxea Open-World dataset [ 1 ]; manufacturing clips from Build AI's Egocentric-10K [ 2 ].
> 
> Despite recent advances in egocentric video understanding, high-quality annotation at scale is still hard. Perceptron Egocentric closes three gaps:
> 
> Subtask labels at a fraction of the cost.

---

Record: https://forck.live/items/31482-introducing-perceptron-egocentric-1-0
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md