Open Set Recognition Project

Ongoing Open Set Recognition project using PyTorch.

For any issue and question, please email ma.xu1@northeastern.edu

Attention: need to be re-constrcuted due to my experimental implementations (especially my methods).

Requirements

For different Algorithms and different datasets, the requirements varies. In general, the basic and must requirements are:

# pytorch 1.4+, torchvision 0.7.0 +
pip3 install torch torchvision
# sklearn
pip3 install -U scikit-learn
# numpy
pip3 install numpy
# scikit-learn-0.23.2
pip3 install -U sklearn

For OpenMax:

pip3 install libmr

For plotting MNIST:

pip3 install imageio
pip3 install tqdm

Supporting

• DataSet
  • CIFAR-100 (done)
  • CIFAR-10 (todo)
  • MNIST (Done)
  • ImageNet (todo)
• Algorithms
  • SoftMax (done)
  • SoftMax with threshold (done)
  • OpenMax(CVPR2016) (done)
  • OLTR (CVPR2019) (done)
  • Center Loss (ECCV2016) (Done)
  • More ...
• Evaluations
  • Accuracy
  • F1-measure
  • More ...

Have a try

Click go link to the related method/dataset and have a try.

		CIFAR-100	CIFAR-10	MNIST	ImageNet
OpenMax	[ReadME]	go
OLTR	[ReadME]	go
CenterLoss	[ReadME]	go