PytorchFastDraw

GPU-accelerated drawing utilities using PyTorch for fast, parallel rendering operations.

Overview

PytorchFastDraw provides a Canvas class that leverages PyTorch tensors and CUDA acceleration to perform high-performance drawing operations. This is particularly useful for applications requiring real-time rendering of many primitives simultaneously.

Features

• GPU Acceleration: All drawing operations run on CUDA-enabled GPUs
• Batch Operations: Draw thousands of points and lines in parallel
• Gradient Lines: Support for color gradient line rendering
• OpenCV Integration: Easy display and export via OpenCV

Requirements

• Python 3.x
• PyTorch (with CUDA support recommended)
• OpenCV (cv2)

Installation

Install torch with CUDA support following the instructions at PyTorch.org, then install the required Python packages:

pip install opencv-python

Usage

from torch_draw import Canvas

# Create a canvas (800x600 pixels on GPU)
canvas = Canvas(800, 600, device='cuda')

# Clear with a color
canvas.clear(torch.tensor([0, 0, 0], device='cuda'))

# Draw points
positions = torch.tensor([[100, 100], [200, 200]], device='cuda')
color = torch.tensor([255, 0, 0], dtype=torch.uint8, device='cuda')
canvas.draw(positions, color)

# Draw two lines (x0, y0, x1, y1)
lines = torch.tensor([[0, 0, 100, 100], [50, 0, 50, 100]], device='cuda')
canvas.drawLine(lines, color)

# Draw gradient lines
canvas.drawLineGradient(
    lines,
    col_start=torch.tensor([255, 0, 0], dtype=torch.uint8, device='cuda'),
    col_end=torch.tensor([0, 0, 255], dtype=torch.uint8, device='cuda')
)

# Display the result
canvas.display()

License

MIT License - Copyright (c) 2025 Filip Optołowicz, University of Wrocław