Lightner is an open-source object recognition microservice. This software rapidly identifies objects on labeled datasets supplied by the user.
Lightner applies a single neural network to a full motion video image. The software divides the image into regions and predicts bounding boxes and probabilities for each region. Each bounding box is weighted by the predicted probability. Lightner examines the whole image, in testing, for global context to ensure informed probability.
Lightner analytical data is provided by frame, in order to observe the code. This allows the user the ability to act when needed without video. The images can be routed to entity extraction or facial recognition.
Lightner is compatible with existing ISR video systems and sensors. It can be used on a low-power, unattended system (internet of things – IOT) and be deployed for on-sensor processing.
This product is available as a bolt-on microservice capable of processing multiple video streams simultaneously. Users can directly ingest JavaScript Object Notation (JSON) format for storing and transporting data from each feed or run analytics using data stored in a graph database.