Timezone: »
We consider the problem of anomaly detection in images, and present a new detection technique. Given a sample of images, all known to belong to a ``normal'' class (e.g., dogs), we show how to train a deep neural model that can detect out-of-distribution images (i.e., non-dog objects). The main idea behind our scheme is to train a multi-class model to discriminate between dozens of geometric transformations applied on all the given images. The auxiliary expertise learned by the model generates feature detectors that effectively identify, at test time, anomalous images based on the softmax activation statistics of the model when applied on transformed images. We present extensive experiments using the proposed detector, which indicate that our algorithm improves state-of-the-art methods by a wide margin.
Author Information
Izhak Golan (Technion)
Ran El-Yaniv (Technion)
More from the Same Authors
-
2022 Poster: TransBoost: Improving the Best ImageNet Performance using Deep Transduction »
Omer Belhasin · Guy Bar-Shalom · Ran El-Yaniv -
2021 Poster: Disrupting Deep Uncertainty Estimation Without Harming Accuracy »
Ido Galil · Ran El-Yaniv -
2019 Poster: Deep Active Learning with a Neural Architecture Search »
Yonatan Geifman · Ran El-Yaniv -
2017 Poster: Multi-Objective Non-parametric Sequential Prediction »
Guy Uziel · Ran El-Yaniv -
2017 Poster: Selective Classification for Deep Neural Networks »
Yonatan Geifman · Ran El-Yaniv -
2012 Poster: Pointwise Tracking the Optimal Regression Function »
Yair Wiener · Ran El-Yaniv -
2012 Spotlight: Pointwise Tracking the Optimal Regression Function »
Yair Wiener · Ran El-Yaniv -
2011 Poster: Selective Prediction of Financial Trends with Hidden Markov Models »
Dmitry Pidan · Ran El-Yaniv -
2011 Poster: Agnostic Selective Classification »
Yair Wiener · Ran El-Yaniv -
2006 Poster: Optimal Single-Class Classification Strategies »
Ran El-Yaniv · Mordechai Nisenson