Adaptable robot vision, built for the factory floor
Lentil automates challenging tasks with minimal engineering time and low cycle times. Its flexibility makes it suitable even for high-mix production.
Automate challenging tasks
- Flexible or deformable objects
- Transparent objects
- Metal on metal
- High precision
- Any part size
Built for production
- Speed: inference in fractions of a second; robot can run at any speed
- Local training; no internet needed
- Hardware agnostic
- Robust to rust, dirt, and part variation
Deploy in minutes
- Training pipeline in minutes; 30 min from hardware setup to robot executing task
- Immediately robust to lighting changes, even for transparent and reflective objects
Capabilities
Robot motion
- Object detection
- Pose estimation
- Visual servoing
Classification
- Detect and classify different objects
- Assembly verification and rudimentary quality inspection
- Robot skill selection
Advanced AI, on any standard hardware
Lentil is pure software. It runs on hardware that meets your needs — components you already own, or purchase from or in consultation with us.
Vision software, licensed to fit
Annual or perpetual license covering models, training tooling, and updates — per cell, in batches, or unlimited for large deployments. Unlimited tasks and re-training, with onboarding included.
Hardware-agnostic
A standard industrial robot, an RGB camera, and an edge compute device. No special lighting, fixtures, or rigs. Add as many cameras as you want at a very low cost.
Lightweight models, locally trained
Data-efficient AI models that run on local edge compute — with a patent-pending training method.
Lightweight, on-device models
Runs locally on edge compute. No cloud, no round-trips, no data leaving the factory.
Built around factory reality
All our tech is shaped by real manufacturing constraints, not generic AI benchmarks. R&D-focused, with one patent pending and more on the way.
Learns from few examples
Production performance from a fraction of the data — no collection campaign required.