This is an assignment from the course of Advanced Computer Architecture from the Integrated Masters in Computer and Telematics Engineering. It consists on an implementation of an image-matching method using parallel computing model CUDA. In this work, all algorithms that were previously implemented in pure C++ were migrated to CUDA implementations, taking advantage of the GPU's huge processing power to largely improve the original approach's efficiency on image-matching, enabling the application of this method in the computer vision area.
- CUDA
- C++
- Parallel computing
- Semi-global Matching
The entire solution was developed by Rui Espinha Ribeiro (Espinha) and David Silva (dmpasilva).