> ## Documentation Index
> Fetch the complete documentation index at: https://docs.datamarkin.com/llms.txt
> Use this file to discover all available pages before exploring further.

# PixelFlow

> Extreme simplicity. Extreme performance. Complete reusability. Open source computer vision that just works.

# Computer Vision Without the Complexity

PixelFlow transforms how you work with computer vision. One library, any framework, maximum performance. Built for developers who need CV to work reliably across different models and environments.

<CardGroup cols={3}>
  <Card title="Extreme Simplicity" icon="bolt">
    One function does one thing perfectly. No complex abstractions or steep learning curves.
  </Card>

  <Card title="Extreme Performance" icon="rocket">
    OpenCV-optimized drawing primitives built for production-speed processing.
  </Card>

  <Card title="Complete Reusability" icon="arrows-rotate">
    Write once, use with YOLO, Detectron2, MediaPipe, or any CV framework.
  </Card>
</CardGroup>

## The Simplest Computer Vision API

Here's how simple it is:

<CodeGroup>
  ```python YOLO + PixelFlow theme={null}
  import pixelflow as pf
  from ultralytics import YOLO

  # Any YOLO model works
  model = YOLO('yolov8n.pt')
  results = model('image.jpg')

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

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

  ```python Detectron2 + PixelFlow theme={null}
  import pixelflow as pf
  from detectron2 import model_zoo
  from detectron2.engine import DefaultPredictor

  # Any Detectron2 model works
  cfg = get_cfg()
  cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml"))
  predictor = DefaultPredictor(cfg)

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

  ```python MediaPipe + PixelFlow theme={null}
  import pixelflow as pf
  import mediapipe as mp

  # Any MediaPipe solution
  mp_pose = mp.solutions.pose.Pose()
  results = mp_pose.process(image)

  # Unified PixelFlow interface
  detections = pf.from_mediapipe(results, image.shape)
  annotated = pf.annotate.keypoint(image, detections)
  ```
</CodeGroup>

<Card title="Get Started Now" icon="rocket" href="quickstart" horizontal>
  Complete installation and first application in under 5 minutes
</Card>

## What You Can Build

PixelFlow works across industries - from retail analytics to manufacturing quality control, security monitoring to healthcare imaging. The unified interface means your code works regardless of the underlying model.

<CardGroup cols={2}>
  <Card title="Object Detection & Tracking" icon="crosshairs">
    Multi-object tracking with ByteTracker, zone-based filtering, and crossing detection for comprehensive scene analysis.
  </Card>

  <Card title="Visual Annotations" icon="paintbrush">
    Professional-grade annotations including boxes, masks, keypoints, blur effects, and custom overlays.
  </Card>

  <Card title="Spatial Analytics" icon="map">
    Zone management, line crossing detection, and spatial filtering for location-based insights.
  </Card>

  <Card title="Performance Tools" icon="chart-line">
    Built-in timing, FPS counters, and memory-efficient video processing for production deployments.
  </Card>
</CardGroup>

## Framework Compatibility

<Note>
  PixelFlow provides unified converters for all major computer vision frameworks, eliminating vendor lock-in and enabling framework mixing within the same project.
</Note>

**Supported Frameworks:**

* **YOLO/Ultralytics** - Real-time object detection
* **Detectron2** - Research-grade instance segmentation
* **MediaPipe** - Hand/pose/face detection
* **MMDetection** - Advanced detection models
* **HuggingFace Transformers** - Vision transformers
* **Custom Models** - Build your own converters

## Quick Navigation

<CardGroup cols={2}>
  <Card title="Detections" icon="target" href="detections/detections">
    Convert any model output to PixelFlow's unified format
  </Card>

  <Card title="Annotations" icon="pen" href="annotators/introduction">
    Professional visual overlays and effects
  </Card>

  <Card title="Tracking" icon="route" href="tracker/bytetrack">
    Multi-object tracking across video frames
  </Card>

  <Card title="Zones" icon="draw-polygon" href="zones">
    Spatial filtering and region management
  </Card>
</CardGroup>

<Tip>
  **New to computer vision?** Start with the [Quickstart Guide](quickstart) for a complete walkthrough from installation to your first working application.
</Tip>