Data obtained from here, specifically the living_room2 RGB-D images in TUM format (without noise), poses also in the TUM format.
psdf/sdf.py contains all the helper sdf methods, some of which aren't being used, looking through the code should make it pretty obvious what's being done. psdf/helpers.py contains one method that compares tsdf and sdf because I was confused and wanted to see if there was a difference, because I couldn't see it visually, turns out there is, I did not write that function, that was GPT-4. psdf/visualizers.py contains visualizer methods for the things we generate, again some methods are no longer in use, it's kind of messed up. psdf/data_loader.py is how I load data, yes the preprocess method does nothing, but that's because the current data I'm using is clean as a whistle, 0 noise to mess with. main.py is where everything is combined.
- Process Data
- Pointcloud from Depth
- SDF
- TSDF
- ESDF (half done?)
- Elevation Map (should be easy enough)
- Bayesian Updating
- Do it live
- Better scene reconstruction