Optimize AI Data Collection at the Edge
LightlyEdge is a smart data selection SDK designed for edge devices, perfect for automotive, robotics, video analytics, and visual inspection industries. Identify and utilize high-signal data in real-time, reducing data transfer and storage costs while optimizing your machine learning models.
Features that make LightlyEdge unique!
Find relevant frames using text or image queries. Our advanced vision language embeddings make searches intuitive and precise.
Set custom triggers to automate data collection based on diversity and semantic similarity, optimizing your training datasets.
Understand your data with comprehensive analytics, showing the frequency and distribution of specific categories.
How LightlyEdge works
With the LightlyEdge SDK running directly on your data collection devices only the frames that match your individual criteria are stored, improving speed and reducing the need for data storage significantly.
Lightweight
Less than 200 MBytes footprint, ideal
for low-power applications.
Blazing fast
Runs >10 fps on embedded CPUs, >100 fps on embedded GPUs. Reported numbers are with batch size 1.
Foundation Model at your Edge
Search images using text inputs or reference images, identifying edge cases without new model deployment.
Detect domain shifts in real-time to ensure your models remain reliable.
Get LightlyEdge
Search images using text inputs or reference images, identifying edge cases without new model deployment.
Compatible with Linux, Windows, MacOS, x86, ARM.
