CEAL: Cost-Effective Active Learning for Deep Image Classification

Unofficial Pytorch Implementation of CEAL

• This is not an official implementation of CEAL!
• I've only tested this code with CIFAR10 and modified CUB200 dataset because of limitation of time and GPU resources I've got.
• Feel free to leave me an issue or make a pull request!

Files

• CEAL.ipynb: CEAL on CIFAR10 dataset
• CEAL_CUB.ipynb: CEAL on modified CUB200 dataset
• model.py: defines model

modified CUB200

• Because of limited GPU resource, I modified original CUB200 dataset with 200 class into 20 classes.
• First, sort the data by class name. (A-Z)
• And select the top 20 classes for training and testing.
• I used train/test split provided by TensorFlow.
  • You can find it on kaggle

Original Paper

Wang, Keze, et al. "Cost-effective active learning for deep image classification." IEEE Transactions on Circuits and Systems for Video Technology 27.12 (2016): 2591-2600.