# Active Learning algorithms for classification

Uses active learning for the one of the given task in paper "Learning Loss for Active Learning" i.e., classification. There is folder named "Active Learning", where I placed all the code for CIFAR-10 and MNIST training.

## Used open-source model

* [Classification on CIFAR-10 using Resnet-18 model](https://https://github.com/kuangliu/pytorch-cifar)
* [TowardsDataScience-Active Learning for classification](https://https://towardsdatascience.com/active-learning-for-object-detection-and-human-pose-estimation-b38678f46686)
* [Learning loss for active learning-Pytorch](https://https://github.com/seominseok0429/Learning-Loss-for-Active-Learning-Pytorch)
## Running the code

Just upload the whole folder on your drive and follow these steps:
### For running with random selection.
!python3 cifar_resnet_clasification.py #from classification folder.
### For running with loss prediction active learning.
!python3 cifar_resnet_clasification.py --use_loss_prediction_al=True

## Same steps for MNIST datasets just write **"mnist_resnet_classification.py"**