Develop a Hierarchical KNN

For a project, I need you to implement a hierarchical K-NN based on dino features on Etehec dataset. Dino is a project of facebook which is based on self supervised vision transforms. You can find the repo from link : [login to view URL] In this repo there is a python file called [login to view URL] . Wİth this file you can basically train a K-nn with dino features without any change. There is a variable called “distance” that is the main variable for entire k-nn (think like a loss function). We will use the same distance metric for creating a hierarchy. There is a paper called “Hierarchical Image Classification using Entailment Cone Embeddings” and they are using ETHEC dataset during their work. Paper Link : [login to view URL] Dataset Link : [login to view URL] In this paper they are used some different hierarchy methods for CNNs that are PLC, M-PLC and etc.(4 different approach) you can find the approaches on page 11 (7. Appendix). You need to implement same hierarchy approaches with k-nn( with dino features/distance ) instead of CNN. During all these different hierarchy implementations you need to provide training results, graphs and all other metrics with a detailed report. YOu do not need to write entire code from scratch or train any CNN. you just need to modify the dino eval-knn file and use their pretrained network to extract features. You just need to write 4 different hierarchy approach on knn.