Image Matching and Retrival

• Dema Ushchapovskyy

• March, 2019

University deep learning coursework, which involved matching and retrieving noisy patches which have position and illumination transformations.

The task involved evaluating and testing a baseline system for the problem and then coming up with improvements to the system. The final approach was to build an autoencoder which denoises the patches and then a second network (ResNet), which matches and retrieves patches.

Final grade: 1st