streamware

Count Objects Example

Count people, vehicles, or other objects using YOLO.

Quick Start

# Count people
sq watch "count people"
sq watch "ile osób"

# Count cars
sq watch "count cars"
sq watch "ile samochodów"

# Count animals
sq watch "count animals"

Features

• YOLO-based counting: Fast and accurate
• Change notifications: “Count increased to 3”
• No LLM needed: Pure computer vision

Output Messages

Event	Message
Initial count	“3 persons visible”
Increase	“Person count increased to 4”
Decrease	“Person count decreased to 2”
Zero	“No persons visible”

Configuration

# .env settings for count mode
SQ_STREAM_FPS=1.0
SQ_YOLO_SKIP_LLM_THRESHOLD=0.2  # Always use YOLO
SQ_USE_GUARDER=false

Python API

from streamware.intent import parse_intent

intent = parse_intent("count people")
print(f"Action: {intent.action}")  # "count"
print(f"Target: {intent.target}")  # "person"

Advanced: Multi-class Counting

from streamware.yolo_detector import YOLODetector

detector = YOLODetector(confidence_threshold=0.3)
results = detector.detect("/path/to/frame.jpg")

# Count by class
counts = {}
for det in results:
    counts[det.class_name] = counts.get(det.class_name, 0) + 1

print(counts)  # {'person': 3, 'car': 2}

This site is open source. Improve this page.