Project Description

This project overlays an rgba image on an rgb image using alpha blending. The alpha blending is implemented as a cuda kernel. For information on what alpha blending is and why it is useful, click here.

CUDAatScaleForTheEnterpriseCourseProject Quickstart

To run this, you must have access to an nvidia gpu and software. As such, you may wish to use a lab environment from the "CUDA at Scale for the Enterprise" Coursera course to execute it. If you have not already, clone this repository, cd to its base directory, and initialize git submodules:

git clone https://github.com/TannerFirl/CUDAatScaleForTheEnterpriseCourseProject.git
cd CUDAatScaleForTheEnterpriseCourseProject
git submodule --init --recursive

Overlay included sample images (test setup)

git submodule --init --recursive
make clean build run

output.png will contain the sample images overlayed using alpha blending.

Overlay custom images

NOTE: the watermark MUST be an RGBA transparent .png file. Not all .png files are RGBA transparent. I have had luck using this tool to convert images to RGBA transparent .png files. Select Depth color: 32 (True color, RGBA, transparent) and download the converted image. For more information regarding types of .png files, click here.

git submodule --init --recursive
make clean build
./bin/watermark <8bit-rgb-img-filename.png> <8bit-rgba-watermark-filename.png> [output.png]

Code Organization

bin/ This folder holds all binary/executable code that is built automatically or manually.

data/ This folder holds all example data in any format.

src/ The source code for this project.

README.md The description of the project so that anyone cloning or deciding if they want to clone this repository can understand its purpose to help with their decision.

Makefile Script for building this project's code

third-party/cuda-samples cuda's samples repository, which is a submodule of this repository.