For our machine learning class, we decided to do Bird Classification. We used two sources of data, with one containing ‘cleaned’, easy-to-classify images, and the other containing mostly ‘unclean’ images with more noise. We also compared types of models, using convolutional neural networks and semi-supervised learning and seeing how accuracy changed.
The full report and results can be found here
Note: the formatting was done according to the ICLR 2023 conference template, as specified by our professor.