This repository contains CycleGAN (Zhu et al., 2017) pytorch code for day-to-night domain tranfer of frames captured in driving-related context. Specifically, models were trained and tested on scenes element of the Berkeley Deep Drive data set (Yu et al., 2018).
Linux or OSX. Python Python 3. CPU or NVIDIA GPU + CUDA CuDNN.
Our original work used the Berkeley Deep Drive data set for training and testing of the CycleGAN model for day-to-night domain transfer. Please refer to Yu et al., 2018, BDD100K: A Diverse Driving Video Database with Scalable Annotation Tooling, ArXiv:1805.04687.
Code is inspired by pytorch-CycleGAN-and-pix2pix by Zhu et al.