Active-Learning-for-Image-Classification

This repository is the unofficial implementation of Active Learning baseline algorithms for image classification.

Requirements

To install requirements:

pip install -r requirements.txt

Training

To train the model(s) in the paper, run this command:

python main.py -AL random -D CIFAR10 --data_dir <path_to_data> --batch_size 128 --lr 0.1 --lrscheduler multistep --milestone 160 --num_epoch 200 --resize 32 --init kaiming

-AL: implemented AL method
- fixed: AL with fixed indices
- random: random selection
- coreset: core set selection [paper]
- vaal: VAAL [paper]
  - num_epoch_vaal (default 100)
- learningloss: learning loss [paper]
  - subset_size (default 10000)
  - epoch_loss (default 120)
  - margin (default 1.0)
  - weight (default 1.0)
- ws: weight decay scheduling [paper]
  - ws_sampling_type
- badge: BADGE selection [paper]
- seqgcn: sequential GCN [paper]
  - subset_size (default 10000)
  - lambda_loss (default 1.2)
  - s_margin (default 0.1)
  - hidden_units (default 128)
  - dropout_rate (default 0.3)
  - lr_gcn (default 0.001)
- tavaal: TA-VAAL[paper]
  - num_epoch_vaal (default 100)
  - weight (default 1.0)
  - subset_size (default 10000)
- bait: BAIT [paper]
- alfamix: ALFA-Mix [paper]
- gradnorm: AL GradNorm [paper]
  - subset_size (default 10000)

Results

Our model achieves the following performance on:

Image Classification on CIFAR10

Model name \ Accuracy over labeled set	600	800	1000
Ours	.	.	.

📋 Include a table of results from your paper, and link back to the leaderboard for clarity and context. If your main result is a figure, include that figure and link to the command or notebook to reproduce it.

Contributing

MIT License

📋 Pick a licence and describe how to contribute to your code repository.

euphoria0-0/Active-Learning-for-Image-Classification