Deep Learning Based Wind Speed Simulation

This work was featured by KPFui at the NVIDIA GTC 2020 conference.

I finetuned the pix2pix GAN on a custom dataset of building facades and created an API on AWS Sagemaker that is then used in a Grasshopper plugin. The model returns back an image with a wind speed value at each pixel, coded from red (0 m/s) to green (120m/s).

Skills: Deep Learning, Computer Vision, AWS

Tools: Python, Pytorch, OpenCV