Optimize open source SFM library -- repost

Hi, I need someone to optimize an open source SFM library to work with certain types of image sets that we have. You can use any open SLAM or SFM library to get the required output, which is to have image matching and localisation. You must have experience in computer vision, SLAM or SFM to do this. Please don't apply if you don't have experience with these technologies. The primary outcomes of this project would be - The SFM algorithm can easily work with our images - The algorithm creates a detailed point cloud, even with very few images - Generating a 3D layouts from the point cloud + images - The algorithm must detect walls and corners within an indoor space, this information must be fed back into a JSON file. - UI with ability to manually add matching points and tracks in case no match is found between images