# 2021 DIAL / Image classification with Semi Supervised Learning

![description_1](https://i.imgur.com/fl86beG.png)
- Build a machine learning model for the image classification with few labeled data where most of the data are unlabeled. Leverage ‘large unlabeled data’ to improve the model performance which originally has ‘small sized labeled data’ only.


Reference :
https://paperswithcode.com/task/semi-supervised-image-classification
https://amitness.com/2020/07/semi-supervised-learning/
https://blog.est.ai/2020/11/ssl/


### 1. **Baseline(100%)**
The score results from trained model with 40,551 labeled dataset. (whole train dataset)
Model : CNN in model.py file

The baseline uses hyper-parameters as below.
![params](https://i.imgur.com/gewatbq.png)