Get Started in Under 5 Minutes

This guide gets you from zero to running computer vision applications with PixelFlow. You’ll install the library, run your first detection, and see the power of unified CV workflows.

Installation

pip install
Development

pip install pixelflow

git clone https://github.com/datamarkin/pixelflow
cd pixelflow
pip install -e .

Your First PixelFlow Application

Here’s how simple it is to get professional computer vision results:

import pixelflow as pf
import cv2
from ultralytics import YOLO

# Load any YOLO model
model = YOLO('yolov8n.pt')  # Downloads automatically first time

# Load your image
image = cv2.imread('your_image.jpg')

# Run detection
results = model(image)

# Convert to PixelFlow format
detections = pf.from_ultralytics(results[0])

# Professional annotations in one line
annotated = pf.annotate.box(image, detections)

# Display or save result
cv2.imshow('PixelFlow Result', annotated)
cv2.waitKey(0)

import pixelflow as pf
import cv2
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg

# Setup Detectron2 model
cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml"))
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml")
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
predictor = DefaultPredictor(cfg)

# Load your image  
image = cv2.imread('your_image.jpg')

# Run detection
outputs = predictor(image)

# Same PixelFlow workflow
detections = pf.from_detectron2(outputs)
annotated = pf.annotate.box(image, detections)

cv2.imshow('PixelFlow Result', annotated)
cv2.waitKey(0)

import pixelflow as pf
import cv2
import mediapipe as mp

# Setup MediaPipe
mp_pose = mp.solutions.pose.Pose()

# Load your image
image = cv2.imread('your_image.jpg')
rgb_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Run pose detection
results = mp_pose.process(rgb_image)

# Unified PixelFlow interface
detections = pf.from_mediapipe(results, image.shape)
annotated = pf.annotate.keypoint(image, detections)

cv2.imshow('PixelFlow Result', annotated)
cv2.waitKey(0)

All examples use the same PixelFlow workflow: Model Output � Convert � Annotate. This pattern works across every supported framework.

Advanced Features in 3 More Lines

Once you have basic detection working, PixelFlow’s advanced features are just as simple:

Object Tracking

# Add multi-object tracking
from pixelflow.tracker import ByteTracker

tracker = ByteTracker()
tracked_detections = tracker.update(detections, image)
annotated = pf.annotate.box(image, tracked_detections, show_ids=True)

Zone-Based Filtering

# Filter detections by spatial zones
from pixelflow.zones import Zones

zones = Zones.from_polygons([[(100, 100), (400, 100), (400, 300), (100, 300)]])
filtered_detections = detections.filter_by_zones(zones)
annotated = pf.annotate.zones(image, zones, detections=filtered_detections)

Privacy Protection

# Blur faces for privacy compliance
person_detections = detections.filter_by_class([0])  # Person class = 0 in COCO
privacy_safe = pf.annotate.blur(image, person_detections)

Complete Working Example

Here’s a full script that demonstrates PixelFlow’s power:

complete_example.py

import pixelflow as pf
import cv2
from ultralytics import YOLO
from pixelflow.tracker import ByteTracker
from pixelflow.zones import Zones

# Setup
model = YOLO('yolov8n.pt')
tracker = ByteTracker()

# Define a zone (rectangle from top-left to bottom-right)
zone_polygon = [(200, 200), (600, 200), (600, 400), (200, 400)]
zones = Zones.from_polygons([zone_polygon])

# Process image
image = cv2.imread('busy_street.jpg')
results = model(image)

# PixelFlow pipeline
detections = pf.from_ultralytics(results[0])
tracked_detections = tracker.update(detections, image)
zone_detections = tracked_detections.filter_by_zones(zones)

# Professional visualization
annotated = image.copy()
annotated = pf.annotate.zones(annotated, zones, alpha=0.3)
annotated = pf.annotate.box(annotated, zone_detections, show_ids=True)
annotated = pf.annotate.label(annotated, zone_detections, show_confidence=True)

# Display results
cv2.imshow('PixelFlow Complete Example', annotated)
cv2.waitKey(0)
cv2.destroyAllWindows()

print(f"Detected {len(zone_detections)} objects in the zone")

Next Steps

You’re now ready to build production computer vision applications! Here’s where to go next:

Detections

Master the unified detection format that works with any model

Annotations

Explore all 20+ professional annotation functions

Object Tracking

Add multi-object tracking to your applications

Spatial Analytics

Use zones and crossings for location-based insights

Pro Tip: PixelFlow’s modular design means you can use any component independently. Start with basic detection and annotations, then add tracking, zones, and advanced features as needed.

Getting Started

Detections

Annotators

Zones

Object Tracking

Processing Tools

Analytics

Quickstart

Get Started in Under 5 Minutes

Installation

Your First PixelFlow Application

Advanced Features in 3 More Lines

Object Tracking

Zone-Based Filtering

Privacy Protection

Complete Working Example

Next Steps

Detections

Annotations

Object Tracking

Spatial Analytics

​Get Started in Under 5 Minutes

​Installation

​Your First PixelFlow Application

​Advanced Features in 3 More Lines

​Object Tracking

​Zone-Based Filtering

​Privacy Protection

​Complete Working Example

​Next Steps

Detections

Annotations

Object Tracking

Spatial Analytics

Get Started in Under 5 Minutes

Installation

Your First PixelFlow Application

Advanced Features in 3 More Lines

Object Tracking

Zone-Based Filtering

Privacy Protection

Complete Working Example

Next Steps