|
--- |
|
license: mit |
|
datasets: |
|
- alkzar90/NIH-Chest-X-ray-dataset |
|
language: |
|
- en |
|
metrics: |
|
- f1 |
|
- accuracy |
|
pipeline_tag: image-classification |
|
tags: |
|
- Few-Shot Learning |
|
- medical |
|
- Computer Vision |
|
- Image Classification |
|
--- |
|
# Few-shot Learning Using Random Subspace |
|
|
|
## Overview |
|
This repository contains the code for our work on few-shot learning for chest X-ray images. Our approach is detailed in our paper, which can be accessed [here](https://openreview.net/pdf?id=AF97JZpgPe). |
|
|
|
For a quick overview of our project, visit our [website](https://few-shot-learning-on-chest-x-ray.github.io/Project-Page/). |
|
|
|
## Project Summary |
|
Our project presents a novel method for few-shot learning, specifically tailored for the analysis of chest X-ray (CXR) images. The key features of our method include: |
|
|
|
- **Efficiency**: Our approach is nearly 1.8 times faster than the traditional t-SVD method for subspace decomposition. |
|
- **Effective Clustering**: The method ensures the creation of well-separated clusters of training data in discriminative subspaces. |
|
- **Promising Results**: We have tested our method on large-scale CXR datasets, yielding encouraging outcomes. |
|
|
|
|
|
## Contact |
|
Reach out to the authors [details provided in the project page] |
