zdraw 0.1.3.6

A high-performance drawing library for AI vision tasks, bounding boxes, and keypoints.

ZDraw

ZDraw is a powerful Python library for drawing custom bounding boxes with rounded corners on images using OpenCV. It features advanced multi-label support, metadata-style panels, PNG icon integration, and intelligent dynamic sizing for professional computer vision applications.

Features

Core Drawing Capabilities

• Rounded rectangles with dynamic corner radius and border thickness
• Universal shape drawing (lines, triangles, rectangles, polygons)
• Keypoint visualization with skeleton connections
• Smart color management with automatic class-to-color mapping

Advanced Label System (ZDrawRectCustom)

• Multi-label support - Display multiple labels per bounding box
• Metadata-style panels - Unified background for all labels (like professional CV applications)
• PNG icon support - Add icons next to labels with full transparency support
• Dynamic positioning - Auto-detection of optimal label placement
• Smart text handling - Automatic truncation for long labels
• Responsive sizing - Font and spacing adapt to frame and bounding box dimensions
• Boundary-aware - Prevents labels from extending outside frame boundaries

Installation

pip install zdraw

Quick Start

Basic Bounding Box with Label

import cv2
from zdraw import ZDraw

# Initialize ZDraw
zdraw = ZDraw()

# Load an image
frame = cv2.imread("image.jpg")
frame = cv2.resize(frame, (800, 600), interpolation=cv2.INTER_AREA)

# Draw simple bounding box with label
x1, y1, x2, y2 = 100, 150, 400, 300
frame = zdraw.ZDrawRect(frame, x1, y1, x2, y2, class_name="person")

Advanced Multi-Label with Metadata Style

# PPE Detection with metadata-style panel
frame = zdraw.ZDrawRectCustom(
    frame, x1, y1, x2, y2,
    main_class="person",
    sub_labels=["helmet", "safety_vest", "steel_boots"],
    metadata_style=True,  # Unified background panel
    label_position="auto"  # Smart positioning
)

# With PNG icons (optional)
icons = {
    "person": "icons/person.png",
    "helmet": "icons/helmet.png",
    "safety_vest": "icons/vest.png"
}

frame = zdraw.ZDrawRectCustom(
    frame, x1, y1, x2, y2,
    main_class="person",
    sub_labels=["helmet", "safety_vest"],
    metadata_style=True,
    icons=icons,  # Optional PNG icons
    label_position="outside_top"
)

Usage Examples

Shape Drawing

# Draw various shapes
line_points = [(50, 50), (200, 50)]
triangle_points = [(250, 100), (350, 100), (300, 200)]
rectangle_points = [(50, 300), (200, 300), (200, 400), (50, 400)]

frame = zdraw.ZDrawShape(frame, line_points, shape="line")
frame = zdraw.ZDrawShape(frame, triangle_points, shape="triangle")
frame = zdraw.ZDrawShape(frame, rectangle_points, shape="rectangle")

### Line Drawing with Endpoints
```python
# Draw a line with endpoints (using unpacked coordinates)
frame = zdraw.ZDrawLine(frame, 50, 50, 200, 200, color=(0, 255, 0), thickness=3, draw_points=True)

# Draw a line using points tuple
pt1 = (300, 50)
pt2 = (450, 200)
frame = zdraw.ZDrawLine(frame, pt1, pt2, color=(0, 0, 255), thickness=2)

### PPE Detection Example

```python
# Real-world PPE monitoring scenario
frame = zdraw.ZDrawRectCustom(
    frame, 100, 100, 300, 400,
    main_class="person",
    sub_labels=["helmet", "vest", "boots"],
    metadata_style=True,
    label_position="auto"
)

# Violation detection
frame = zdraw.ZDrawRectCustom(
    frame, 400, 100, 600, 400,
    main_class="person",
    sub_labels=["violator"],
    metadata_style=True,
    label_position="outside_top"
)

API Reference

Core Methods

ZDraw(class_colors=None)

Initializes the ZDraw object.

Parameters:

• class_colors (dict, optional): Dictionary mapping class names to RGB tuples

ZDrawRect(frame, x1, y1, x2, y2, class_name=None, color=None, return_original_frame=False)

Draws a simple bounding box with optional label.

Parameters:

• frame: Input image frame
• x1, y1, x2, y2: Bounding box coordinates
• class_name: Optional class label
• color: Optional color override
• return_original_frame: Whether to return original frame

ZDrawRectCustom(frame, x1, y1, x2, y2, main_class, sub_labels=None, color=None, label_position="auto", metadata_style=True, icons=None, return_original_frame=False)

Enhanced bounding box with multi-label support and metadata-style panels.

Parameters:

• frame: Input image frame
• x1, y1, x2, y2: Bounding box coordinates
• main_class: Primary class label
• sub_labels: List of additional labels
• color: Optional color override
• label_position: Label positioning ("auto", "inside", "outside_top", "outside_bottom", "outside_left", "outside_right")
• metadata_style: If True, displays unified metadata panel (default: True)
• icons: Dictionary mapping label names to PNG icon file paths (optional)
• return_original_frame: Whether to return original frame

ZDrawShape(frame, points, shape=None, return_original_frame=False)

Draws universal shapes (lines, triangles, rectangles, polygons).

Parameters:

• frame: Input image frame
• points: List of points defining the shape
• shape: Optional shape type for validation
• return_original_frame: Whether to return original frame

ZDrawKeypoints(frame, keypoints, skeleton=None, point_color=(0, 255, 255), line_color=(255, 0, 0), radius=3, thickness=2)

Draws keypoints with optional skeleton connections.

Parameters:

• frame: Input image frame
• keypoints: List of (x, y, visibility) tuples
• skeleton: Optional list of (idx1, idx2) connections
• point_color: Color for keypoints
• line_color: Color for skeleton lines
• thickness: Line thickness
• radius: Keypoint radius

ZDrawLine(frame, *args, color=None, thickness=2, draw_points=False, point_radius=None, return_original_frame=False)

Draws a line with flexible coordinate inputs and optional endpoint circles.

Parameters:

• frame: Input image frame.
• *args: Coordinates defining the line. Can be:
  • 4 arguments: x1, y1, x2, y2
  • 2 arguments: (x1, y1), (x2, y2)
  • 1 argument: [(x1, y1), (x2, y2)]
• color: Color (BGR). Defaults to white.
• thickness: Line thickness.
• draw_points: If True, draws filled circles at endpoints.
• point_radius: Radius of endpoint circles. Defaults to thickness + 3.
• return_original_frame: Whether to return original frame.

Label Positioning Options

• "auto": Automatically chooses the best position based on available space
• "inside": Places labels inside the bounding box
• "outside_top": Places labels above the bounding box
• "outside_bottom": Places labels below the bounding box
• "outside_left": Places labels to the left of the bounding box
• "outside_right": Places labels to the right of the bounding box

Icon Support

Icons should be PNG files with transparency support. The icons parameter accepts a dictionary mapping label names to file paths:

icons = {
    "person": "path/to/person.png",
    "helmet": "path/to/helmet.png",
    "vest": "path/to/vest.png"
}

Dependencies

• Python 3.8 or higher
• OpenCV (opencv-python)
• NumPy

License

This project is licensed under the MIT License. See the LICENSE file for details.