Learn computer graphics by writing GPU shaders!

Overview

Graphics Workshop

This repo contains a selection of projects designed to help you learn the basics of computer graphics. We'll be writing shaders to render interactive two-dimensional and three-dimensional scenes.

Each of these projects aims to introduce an important graphics technique that is extensively used in real-world applications. The code is designed to run in real time on modern GPUs, without requiring any extra software. Feel free to take creative liberties to produce a result that you find aesthetically pleasing!


Contents


Getting started

This workshop covers fragment shaders (GLSL), procedural texture generation, rasterization, lighting calculations, and real-time ray tracing.

All of the projects will be developed in a harness that runs the graphics code in a web browser, using a standard technology called WebGL. This allows us to use modern web development tools to iterate quickly and easily share your work with others. However, you will not need to write JavaScript code yourself, as all of the wrapper code has been provided for you.

To get started, make sure that you have Node.js v14 and NPM installed. Fork this repository, and then clone it to your computer. Then, run the following commands:

$ npm install
...
added 16 packages from 57 contributors and audited 16 packages in 1.459s

3 packages are looking for funding
  run `npm fund` for details

found 0 vulnerabilities


$ npm run dev

  vite v2.1.5 dev server running at:

  > Local:    http://localhost:3000/
  > Network:  http://10.250.14.217:3000/

  ready in 555ms.

Explanation: The first command installs dependencies in the node_modules/ folder, and the second command starts the development server. You can now visit http://localhost:3000/ in your browser. Whenever you change a file and save it, the website will automatically update.


Projects

Here is an introduction to each of the projects.

If you are new to graphics or shading languages, I would recommend trying the "quilt patterns" project first. You should also complete "rasterization and shading" before starting the "stylized rendering" and "ray tracing" projects.

Quilt patterns

This project shows how you can use GPU shaders to render interesting, procedural patterns on a 2D grid. It also serves as a hand-on introduction to the wonderful yet frightening world of GPU shaders.

Before starting, make sure that you are able to open the development server. You should see an oscillating color gradient in your web browser.

  1. Look at the provided skeleton code. This is known as a fragment shader, and it runs on the GPU. The fragment shader is run on every pixel of the screen, and the pixel number is passed into the code as gl_FragCoord.xy. The code outputs by assigning to gl_FragColor, which is the color at that pixel. It is a vec4 or 4-dimensional vector with red, green, blue, and transparency channels.
  2. Make sure you understand the starter code before continuing. What do the coord and color variables represent? What do the abs() and sin() functions do?
  3. Uncomment the numbered sections of the starter code, one at a time. Observe the change in output on your screen. Make sure that you understand what each block of code is doing before proceeding to the next one! When you're done, you should now see a moving black-and-white grid pattern.
  4. Task 1: Make the quilt pattern more colorful! Modify the code so that instead of generating random greyscale colors with the variable c, it generates random RGB colors instead.
  5. Task 2: Right now, there's just a single pattern that's flown through, but there could be many more. Find a way to use the seed variable, which you can modify using the "Parameters" panel at the top right of the screen, to change the shape of the pattern. (Hint: You'll probably want to adjust the condition if (mod(2.0 * cell.x + cell.y, 5.0) == 1.0) to change the quilt's geometry.)

Feel free to experiment and get creative! For example, add stripes or other shapes besides triangles. If you're looking for inspiration, try doing a Google search for "quilt patterns."

Project skeleton is located in shaders/quilt.frag.glsl.

Procedural landscapes

This project involves generating an organic procedural landscape, similar to what you might see in open world games like Minecraft.

To view in your browser, choose the "project" setting in the controls pane on the top-right and select this project.

  1. To generate natural appearances, we will be using a popular graphics primitive called simplex noise. Feel free to read online about the theory behind this algorithm. The function float snoise(vec2) takes in a 2-vector on the screen and outputs a smooth scalar noise value at that location.
  2. Try changing the "seed" at the top-right of the screen. You should see the rendered output take a new shape, as noise values at different locations are roughly independent.
  3. Uncomment the first block of code, one line at a time. What do you notice visually as higher-frequency noise scales are added to the color? By combining various octaves of noise at different amplitudes, you can change the texture's appearance.
  4. Uncomment the second block of code, and observe how thresholding (in particular, the mix and smoothstep functions) are used to interpolate colors between elevation values.
  5. Finally, uncomment the third block of code to generate land objects.
  6. Task: In addition to elevation, add another parameter such as temperature that is independently sampled from the noise map. Experiment with using temperature to change the shading and indicate map biomes.

Project skeleton is located in shaders/landscape.frag.glsl.

Rasterization and shading

This project renders a 3D triangle mesh using the rasterization method that is very popular in real-time computer graphics. This is the same algorithm that most video games are built upon.

Rasterization works by breaking up a 3D surface into triangles, then drawing each triangle on the screen independently and interpolating variables between them.

Three-dimensional graphics can be much more complex than its two-dimensional counterpart. Unlike in the past projects, the fragment shader will now be evaluated once for each fragment of a triangle, rather than once for every pixel. For this project, we will be focus on shading. See the Scratchapixel article on the Phong illumination model for background reading.

  1. The project starts out by giving you a teapot. You can click and drag on the screen to move your camera position and see different angles of the teapot. This object is stored as a triangle mesh, which is a common format in computer graphics.
  2. Try changing the value of mesh to see your code rendering different shapes besides the teapot. Also, you can play with kd to change the color of the object.
  3. Task: Right now, the starter code has a function illuminate() which takes in the position of a light and returns how much color that light contributes to the current pixel. At the moment, the rendered result looks a little "flat" because the code only implements the diffuse component of the Phong reflectance model. Based on the Wikipedia article, and using the comments as a guide, update the code to also add the Phong specular component. Your task is complete when the teapot looks similar to the image above.

If you're curious to learn more about lighting models, try upgrading your Phong shader to a microfacet model such the Cook-Torrance BRDF. If you do, you may also want to implement sRGB gamma correction.

Project skeleton is located in shaders/shading.frag.glsl.

Stylized rendering

This project explores stylized rendering, also known as non-photorealistic rendering. This is an area of graphics that gives up being faithful to real life, but enables much more creative expression in mimicking expressive styles.

The code for this project is very similar to that of the last rasterization and shading project. However, after doing lighting calculations, we do not output the color immediately. Instead, we discretize it and shade in different styles based on a thresholded value of the luma intensity. Here, we mimic the art style found in comic books.

  1. Task: Once again, fill in your code for specular highlights in the illuminate() function, similar to last time. How does this change the output's appearance as you rotate the camera view? You may need to tweak some parameters to obtain an aesthetically pleasing result.

Project skeleton is located in shaders/contours.frag.glsl.

Ray tracing

This project is a real-time ray tracer implemented as a GPU shader. Ray tracing is the gold standard in photorealistic rendering. Here's how the Academy Award-winning textbook Physically Based Rendering describes it:

pbrt is based on the ray-tracing algorithm. Ray tracing is an elegant technique that has its origins in lens making; Carl Friedrich Gauß traced rays through lenses by hand in the 19th century. Ray-tracing algorithms on computers follow the path of infinitesimal rays of light through the scene until they intersect a surface. This approach gives a simple method for finding the first visible object as seen from any particular position and direction and is the basis for many rendering algorithms.

You might find this non-technical introductory video by Disney to provide helpful context.

  1. To parallelize our ray tracer and run it on a GPU shader, we revert back to running our fragment shader once for every pixel in the screen, like in the 2D graphics examples. However, this time, our shader will do some geometry calculations by shooting rays from the camera for each pixel. The trace() function models this behavior by returning the color corresponding to a given ray.
  2. To model geometry, the starter code has set up a simple scene involving three spheres and one circle forming the checkered floor. This geometry is located in the intersect() function, which computes the first intersection point (represented by a Hit struct) of any ray in the space. Following the pattern shown here, add a fourth sphere to the scene with a different position, color, and radius.
  3. The calcLighting() function uses illuminate(), which contains a simple implementation of a Phong shader, to compute the lighting at a given point by adding up the contributions of two point lights. Modify this code to add a third point light at a different location, with your choice of intensity. (For inspiration, see three-point lighting.)
  4. Task: Implement shadows. In the illuminate() function, before doing lighting calculations, add a conditional statement checks if the ray from the point to the light source is obstructed or not. If it is obstructed, then you should immediately return vec3(0.0) to simulate shadows. [Hint: You will need to use the intersect() function with a new Ray object that you construct, starting from the point pos.]

If you're interested in learning more about ray tracing, check out rpt, which is a physically-based path tracer written in Rust with a lot more features.

Project skeleton is located in shaders/raytracing.frag.glsl.


Workflow

Open your code editor in one window, and keep the website on the side while you're writing your shader code. It's extremely useful to see what happens to the output in real time as you edit your code. Your browser should instantly update (using Vite's hot module replacement feature) whenever you change a file.

The project has a Tweakpane component in the top-right corner of the screen. You can use this to change what project is being displayed and modify parameters for that display. It persists settings when you refresh your browser.

For 3D projects, the main element has an event listener for mouse and touch events, which should allow you to move the camera by clicking and dragging. Data is passed in through the appropriate uniform variables.

When you're done with the workshop, each project should render without errors or visual artifacts, and the outputs should be interesting.


Deployment

This repository comes with a Continuous Integration workflow (through GitHub Actions) that compiles the code on the main branch, then pushes the built HTML to your repository's gh-pages branch. It lets you share a web link that updates whenever you push code changes.

If this isn't enabled already, go to the "Pages" section under the "Settings" tab of your repository and set the deploy target to the gh-pages branch. Then, after the next git push, you will be see your work at https://.github.io/graphics-workshop/.


Debugging tips

If you're stuck debugging OpenGL shaders, try setting gl_FragColor to the result of an intermediate variable and visually examining the output.

If there's a function you're confused about, you can search online through the Khronos documentation. You can also try reading The Book of Shaders, which is an excellent technical resource.


All code is licensed under the MIT license.
Owner
Eric Zhang
Student at Harvard University and competitive programmer, with research in machine learning & programming languages. @scaleapi, @NVIDIA, @HarvardPL
Eric Zhang
[python3.6] 运用tf实现自然场景文字检测,keras/pytorch实现ctpn+crnn+ctc实现不定长场景文字OCR识别

本文基于tensorflow、keras/pytorch实现对自然场景的文字检测及端到端的OCR中文文字识别 update20190706 为解决本项目中对数学公式预测的准确性,做了其他的改进和尝试,效果还不错,https://github.com/xiaofengShi/Image2Katex 希

xiaofeng 2.7k Dec 25, 2022
Detect text blocks and OCR poorly scanned PDFs in bulk. Python module available via pip.

doc2text doc2text extracts higher quality text by fixing common scan errors Developing text corpora can be a massive pain in the butt. Much of the tex

Joe Sutherland 1.3k Jan 04, 2023
Markup for note taking

Subtext: markup for note-taking Subtext is a text-based, block-oriented hypertext format. It is designed with note-taking in mind. It has a simple, pe

Gordon Brander 224 Jan 01, 2023
"Very simple but works well" Computer Vision based ID verification solution provided by LibraX.

ID Verification by LibraX.ai This is the first free Identity verification in the market. LibraX.ai is an identity verification platform for developers

LibraX.ai 46 Dec 06, 2022
Packaged, Pytorch-based, easy to use, cross-platform version of the CRAFT text detector

CRAFT: Character-Region Awareness For Text detection Packaged, Pytorch-based, easy to use, cross-platform version of the CRAFT text detector | Paper |

188 Dec 28, 2022
Code for AAAI 2021 paper: Sequential End-to-end Network for Efficient Person Search

This repository hosts the source code of our paper: [AAAI 2021]Sequential End-to-end Network for Efficient Person Search. SeqNet achieves the state-of

Zj Li 218 Dec 31, 2022
A novel region proposal network for more general object detection ( including scene text detection ).

DeRPN: Taking a further step toward more general object detection DeRPN is a novel region proposal network which concentrates on improving the adaptiv

Deep Learning and Vision Computing Lab, SCUT 151 Dec 12, 2022
Text page dewarping using a "cubic sheet" model

page_dewarp Page dewarping and thresholding using a "cubic sheet" model - see full writeup at https://mzucker.github.io/2016/08/15/page-dewarping.html

Matt Zucker 1.2k Dec 29, 2022
Select range and every time the screen changes, OCR is activated.

ASOCR(Auto Screen OCR) Select range and every time you press Space key, OCR is activated. 範囲を選ぶと、あなたがスペースキーを押すたびに、画面が変わる度にOCRが起動します。 usage1: simple OC

1 Feb 13, 2022
(CVPR 2021) ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection

ST3D Code release for the paper ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection, CVPR 2021 Authors: Jihan Yang*, Shaoshu

CVMI Lab 224 Dec 28, 2022
MONAI Label is a server-client system that facilitates interactive medical image annotation by using AI.

MONAI Label is a server-client system that facilitates interactive medical image annotation by using AI. It is an open-source and easy-to-install ecosystem that can run locally on a machine with one

Project MONAI 344 Dec 23, 2022
零样本学习测评基准,中文版

ZeroCLUE 零样本学习测评基准,中文版 零样本学习是AI识别方法之一。 简单来说就是识别从未见过的数据类别,即训练的分类器不仅仅能够识别出训练集中已有的数据类别, 还可以对于来自未见过的类别的数据进行区分。 这是一个很有用的功能,使得计算机能够具有知识迁移的能力,并无需任何训练数据, 很符合现

CLUE benchmark 27 Dec 10, 2022
TextField: Learning A Deep Direction Field for Irregular Scene Text Detection (TIP 2019)

TextField: Learning A Deep Direction Field for Irregular Scene Text Detection Introduction The code and trained models of: TextField: Learning A Deep

Yukang Wang 101 Dec 12, 2022
一键翻译各类图片内文字

一键翻译各类图片内文字 针对群内、各个图站上大量不太可能会有人去翻译的图片设计,让我这种日语小白能够勉强看懂图片 主要支持日语,不过也能识别汉语和小写英文 支持简单的涂白和嵌字

574 Dec 28, 2022
STEFANN: Scene Text Editor using Font Adaptive Neural Network

STEFANN: Scene Text Editor using Font Adaptive Neural Network @ The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020.

Prasun Roy 208 Dec 11, 2022
EQFace: An implementation of EQFace: A Simple Explicit Quality Network for Face Recognition

EQFace: A Simple Explicit Quality Network for Face Recognition The first face recognition network that generates explicit face quality online.

DeepCam Shenzhen 141 Dec 31, 2022
Line based ATR Engine based on OCRopy

OCR Engine based on OCRopy and Kraken using python3. It is designed to both be easy to use from the command line but also be modular to be integrated

948 Dec 23, 2022
PAGE XML format collection for document image page content and more

PAGE-XML PAGE XML format collection for document image page content and more For an introduction, please see the following publication: http://www.pri

PRImA Research Lab 46 Nov 14, 2022
CVPR 2021 Oral paper "LED2-Net: Monocular 360˚ Layout Estimation via Differentiable Depth Rendering" official PyTorch implementation.

LED2-Net This is PyTorch implementation of our CVPR 2021 Oral paper "LED2-Net: Monocular 360˚ Layout Estimation via Differentiable Depth Rendering". Y

Fu-En Wang 83 Jan 04, 2023
In this project we will be using the live feed coming from the webcam to create a virtual mouse with complete functionalities.

Virtual Mouse Using OpenCV In this project we will be using the live feed coming from the webcam to create a virtual mouse using hand tracking. Projec

Hassan Shahzad 8 Dec 20, 2022