Interpretable Discovery in Large Image Data Sets
This addresses the need for interpretable image analysis in applications like surveillance and science, though it is incremental as it builds on existing CNN and novelty detection techniques.
The paper tackled the problem of detecting novel or anomalous images in large datasets using CNNs, which are accurate but hard to interpret, and introduced a method combining novelty detection with CNN features to achieve rapid discovery with interpretable explanations, applied to ImageNet and planetary science images.
Automated detection of new, interesting, unusual, or anomalous images within large data sets has great value for applications from surveillance (e.g., airport security) to science (observations that don't fit a given theory can lead to new discoveries). Many image data analysis systems are turning to convolutional neural networks (CNNs) to represent image content due to their success in achieving high classification accuracy rates. However, CNN representations are notoriously difficult for humans to interpret. We describe a new strategy that combines novelty detection with CNN image features to achieve rapid discovery with interpretable explanations of novel image content. We applied this technique to familiar images from ImageNet as well as to a scientific image collection from planetary science.