In this project, we study image enhancement to improve the robustness of loop closure process in a SLAM system. We compare a image enhancement technique based on GAN with some model-based methods.
The GAN-based method is based on Tensorflow2 framework. The code is modified from WaterGAN, which is based on Tensorflow1 framework. Many changes on the network structure have been made.
This method were compared with other methods including:
-
DCP
-
UDCP
In order to evaluate the methods, some metrics were leveraged in this project