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).
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
- 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 ...
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 |