linjiatai/umap-based-visualization-for-feature-space-of-self-supervised-learning

Umap-based-visualization-for-feature-space-of-self-supervised-learning

Introduction

You can use this program to observe the feature space of trained checkpoint by self-supervised lerannig.

Requirements

GPU
numpy = 1.22.3
umap-learn = 0.5.3

Usage

You can use the following command to install the umap tool

• Install python dependencies.

pip install numpy==1.22
pip install umap-learn

• Preparation work

(1) You should place the checkpoint of self-supervised learning in checkpoint fold.

(2) You should place a dataset in dataset fold, such as KME (Baidu Netdisk with code wfzk):

dataset

|___KME
    |___ADI
    |___BACK
    |___DEB
    |___LYM
    |___MUC
    |___MUS
    |___NORM
    |___STR
    |___TUM