• Stars
    star
    113
  • Rank 310,115 (Top 7 %)
  • Language
    Jupyter Notebook
  • Created over 3 years ago
  • Updated over 2 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

ICCV2021 - training a post-hoc lightweight GAN-discriminator for open-set recognition

OpenGAN: Open-Set Recognition via Open Data Generation

ICCV 2021 (best paper honorable mention)

alt text

[website] [poster] [slides] [oral presentation] [paper] [PAMI Version 18MB]

Real-world machine learning systems need to analyze novel testing data that differs from the training data. In K-way classification, this is crisply formulated as open-set recognition, core to which is the ability to discriminate open-set data outside the K closed-set classes. Two conceptually elegant ideas for open-set discrimination are: 1) discriminatively learning an open-vs-closed binary discriminator by exploiting some outlier data as the open-set, and 2) unsupervised learning the closed-set data distribution with a GAN and using its discriminator as the open-set likelihood function. However, the former generalizes poorly to diverse open test data due to overfitting to the training outliers, which unlikely exhaustively span the open-world. The latter does not work well, presumably due to the instable training of GANs. Motivated by the above, we propose OpenGAN, which addresses the limitation of each approach by combining them with several technical insights. First, we show that a carefully selected GAN-discriminator on some real outlier data already achieves the state-of-the-art. Second, we augment the available set of real open training examples with adversarially synthesized "fake" data. Third and most importantly, we build the discriminator over the features computed by the closed-world K-way networks. Extensive experiments show that OpenGAN significantly outperforms prior open-set methods.

keywords: out-of-distribution detection, anomaly detection, open-set recognition, novelty detection, density estimation, generative model, discriminative model, adverserial learning, image classification, semantic segmentation.

If you find our model/method/dataset useful, please cite our work (ICCV version on arxiv, PAMI version):

@inproceedings{OpenGAN,
  title={OpenGAN: Open-Set Recognition via Open Data Generation},
  author={Kong, Shu and Ramanan, Deva},
  booktitle={ICCV},
  year={2021}
}

@inproceedings{OpenGAN_PAMI,
  title={OpenGAN: Open-Set Recognition via Open Data Generation},
  author={Kong, Shu and Ramanan, Deva},
  booktitle={IEEE PAMI},
  year={2022}
}

last update: July, 2021

Shu Kong

aimerykong At g-m-a-i-l dot com

More Repositories

1

deepImageAestheticsAnalysis

ECCV2016 - fine-grained photo aesthetics rating with interpretability
MATLAB
296
star
2

Low-Rank-Bilinear-Pooling

CVPR2017 - an ultra-compact bilinear model for fine-grained classification
C++
149
star
3

Recurrent-Pixel-Embedding-for-Instance-Grouping

CVPR2018 - pixel embedding & grouping for structured prediction, e.g., instance segmentation
MATLAB
144
star
4

predictive-filter-flow

Predictive Filter Flow for fully/self-supervised learning on various vision tasks
Jupyter Notebook
140
star
5

Pixel-Attentional-Gating

Pixel Attentional Gating for Parsimonious Per-Pixel Labeling
MATLAB
46
star
6

Recurrent-Scene-Parsing-with-Perspective-Understanding-in-the-loop

CVPR2018 - scene parsing network regulated by geometric prior
MATLAB
37
star
7

Dimensional-Emotion-Analysis-of-Facial-Expression

MATLAB
11
star
8

pollenDetClsSystem

Automated Recognition and Counting of Pollen Grain Species from Field Samples
Jupyter Notebook
8
star
9

deepPollen

PNAS 2020 - a pioneering work that greatly enhances plant ecological & evolutionary research using pollen data -- NSF News
MATLAB
4
star
10

wormSegCountSystem

MATLAB
2
star
11

simpleUI_wormAnnotation

a simple UI for annotating C. elegans in images
MATLAB
2
star
12

PatchMatchingForPollenIdentification

Selecting discriminative patches to match pollen grains for identification.
MATLAB
2
star
13

Identify-Fossil-Pollen-with-Modern-Reference

Match patches with modern pollen grains to identify fossilized grains.
MATLAB
2
star
14

modern_pollen_24wayCls

Joint detection, segmentation and classification with multiplicative architecture.
Jupyter Notebook
2
star
15

Submodular-Exemplar-Selection

The component of submodular exemplar selection for discriminative dictionary to identify pollen grains
MATLAB
2
star
16

MidFea_NSlayer

Learning Mid-Level Features and Modeling Neuron Selectivity for Image Classification
MATLAB
1
star
17

videoSeg

MATLAB
1
star