Vector Graphics Non-photorealistic Rendering

向量圖形非寫實電腦計算繪製

Project Page

This work is inspired by Differentiable Rasterizer for Vector Graphics, using vector-based initial shapes such as strips (unclosed vector paths), blobs (closed vector paths), and circles (vector circles) to create non-photorealistic vector graphics.

本研究受到 Differentiable Rasterizer for Vector Graphics 的啟發，使用彩帶(向量不封閉曲線)、色塊(向量封閉曲線)以及圓形(向量圓形)等向量初始形狀來實作向量圖形的非寫實電腦計算繪製。

Colab GUI Version

To use this code as a tool to generate non-photorealistic vector results, we provide a simple GUI user interface using Google Colab.
Run this code on Google Colab

Install

git submodule update --init --recursive
conda install -y pytorch torchvision -c pytorch
conda install -y numpy
conda install -y scikit-image
conda install -y -c anaconda cmake
conda install -y -c conda-forge ffmpeg
pip install svgwrite
pip install svgpathtools
pip install cssutils
pip install numba
pip install torch-tools
pip install visdom
python setup.py install

Building in debug mode

python setup.py build --debug install

Run

cd apps

Vector Paths

including strips (unclosed vector paths) and blobs (closed vector paths)

painterly_rendering_time.py [-h] [--exp_name EXP_NAME] [--num_paths NUM_PATHS]
                       [--max_width MAX_WIDTH] [--use_lpips_loss]
                       [--num_iter NUM_ITER] [--use_blob]
                       target

e.g., Strip

python painterly_rendering_time.py imgs/fuyue/01.jpg --exp_name test --num_paths 500 --max_width 5.0 --num_iter 200

Blob

python painterly_rendering_time.py imgs/fuyue/01.jpg --exp_name test --num_paths 500 --max_width 5.0 --num_iter 200 --use_blob

Vector Circle

painterly_rendering_circle.py [-h] [--exp_name EXP_NAME] [--num_circles NUM_CIRCLES]
                       [--max_radius MAX_RADIUS] [--max_radius_factor MAX_RADIUS_FACTOR] [--use_lpips_loss]
                       [--num_iter NUM_ITER] [--use_random_position]
                       target

e.g.,

python painterly_rendering_circle.py imgs/fuyue/01.jpg --exp_name test --num_circles 500 --max_radius 5.0 --max_radius_factor 2.0 --num_iter 200 --use_random_position

or to generate circles in fixed position
(number of circles will depend on the size of your input image and the max_radius parameter)

python painterly_rendering_circle.py imgs/fuyue/01.jpg --exp_name test --max_radius 5.0 --max_radius_factor 2.0 --num_iter 200