Feature Representation in Convolutional Neural Networks
This work addresses feature extraction for computer vision tasks, offering incremental improvements by optimizing feature usage in existing models.
The paper tackled the problem of understanding and improving feature representation in Convolutional Neural Networks (CNNs) for image classification, showing that using lower-layer CNN features with Random Forests or SVMs outperforms the original CNN and yields competitive accuracy even with suboptimal CNNs.
Convolutional Neural Networks (CNNs) are powerful models that achieve impressive results for image classification. In addition, pre-trained CNNs are also useful for other computer vision tasks as generic feature extractors. This paper aims to gain insight into the feature aspect of CNN and demonstrate other uses of CNN features. Our results show that CNN feature maps can be used with Random Forests and SVM to yield classification results that outperforms the original CNN. A CNN that is less than optimal (e.g. not fully trained or overfitting) can also extract features for Random Forest/SVM that yield competitive classification accuracy. In contrast to the literature which uses the top-layer activations as feature representation of images for other tasks, using lower-layer features can yield better results for classification.