Drawing

The Painter class allows you to draw either onto an existing image or onto an empty canvas. Simply set it up via:

Python API - How to set up the Painter.

import viren2d

painter = viren2d.Painter()

# Either create an empty canvas:
painter.set_canvas_rgb(height=600, width=800, color='navy-blue')

# Initialize with a given NumPy ndarray `np_image`:
painter.set_canvas_image(image_numpy)

# Or initialize from an image file:
painter.set_canvas_filename('path/to/image.jpg')

C++ API - How to set up the viren2d::Painter.

#include <viren2d/viren2d.h>

auto painter = viren2d::CreatePainter();

// Either create an empty canvas (H=600, W=800):
painter->SetCanvas(600, 800, viren2d::Color(viren2d::NamedColor::NavyBlue));

// Initialize with an existing viren2d::ImageBuffer:
painter->SetCanvas(image_buffer);

// Or initialize from an image file:
painter->SetCanvas("path/to/image.jpg");

Once you’ve set up the painter, you can start drawing:

• Basic Shapes
• Markers & Keypoints
• Text & Text Boxes
• Image Overlays
• Bounding Boxes & Trajectories
• Pose Estimation Results
• What about 3D?
• Camera Geometry

Finally, you simply retrieve the visualization result via get_canvas(). This will yield an ImageBuffer, which can easily be used with common image processing libraries, such as NumPy or OpenCV. The following examples show one option to retrieve your first visualization from the Painter:

Python API - How to retrieve the visualization from a Painter.

# Get the visualization as `viren2d.ImageBuffer`:
img_buffer = painter.get_canvas()

# Convert to NumPy array (avoid unnecessary copy):
import numpy as np
img_np = np.array(img_buffer, copy=False)

# For convenience, the canvas is also exposed as property of the painter:
img_np = np.array(painter.canvas, copy=False)

C++ API - How to retrieve the visualization from a viren2d::Painter.

// Get a deep copy of the painter's canvas (for the following channel swap):
viren2d::ImageBuffer canvas = painter->GetCanvas(true);

// Optionally, if you've set up OpenCV, you can simply create a cv::Mat header
// from the ImageBuffer. Note that OpenCV expects images in BGR(A) format:
#include <opencv2/opencv.hpp>

canvas.SwapChannels(0, 2);
cv::Mat cv_buffer(
    canvas.Height(), canvas.Width(),
    CV_MAKETYPE(CV_8U, canvas.Channels()),
    canvas.MutableData(), canvas.RowStride());

Tip

You can retrieve either a deep copy or a shared view on the Painter’s canvas. If you can ensure that the Painter instance is not destroyed and you don’t concurrently change the canvas, you should prefer the shared view because it avoids unnecessary memory allocation.

Refer to get_canvas() for details.

Note

As demonstrated by the example code above, the viren2d Python API is simply a wrapper of the C++ API. Except for the naming conventions (Snake case in Python vs. camel case in C++), both APIs are highly similar.

Thus, the rest of this tutorial will mostly omit the C++ examples. The corresponding C++ counterpart will instead be mentioned in the API documentation of the Python bindings.