Perceptions

Define what you want to see using natural language queries.

What are Perceptions?

Perceptions are the core concept of Nara Lens. They allow you to define detection criteria using simple, human-readable strings instead of complex model configurations.

Perception Syntax

Perceptions follow a simple action:target format:

detect:person           # Find all people
count:car               # Count cars
find:red vehicle        # Find red vehicles
locate:face looking left # Find faces looking left

Available Actions

detectFind objects with bounding boxes

countCount occurrences of objects

findSearch with natural language descriptions

locateFind objects matching specific criteria

segmentGet pixel-level masks

Examples

# Multiple targets
"detect:person,vehicle,bicycle"

# Natural language
"find:person wearing red shirt"

# Counting
"count:cars in parking lot"

# Complex queries
"locate:face with eyes closed"