Deep Neural Networks Under Stress
This work addresses the need for understanding feature robustness in transfer learning for computer vision applications, but it is incremental as it builds on existing deep learning methods.
The paper tackles the problem of analyzing the resiliency of deep neural network features under compression and perturbations, showing that deep features are more robust than classical approaches, achieving a 98.4% compression rate with only a 0.88% performance loss on Pascal VOC 2007.
In recent years, deep architectures have been used for transfer learning with state-of-the-art performance in many datasets. The properties of their features remain, however, largely unstudied under the transfer perspective. In this work, we present an extensive analysis of the resiliency of feature vectors extracted from deep models, with special focus on the trade-off between performance and compression rate. By introducing perturbations to image descriptions extracted from a deep convolutional neural network, we change their precision and number of dimensions, measuring how it affects the final score. We show that deep features are more robust to these disturbances when compared to classical approaches, achieving a compression rate of 98.4%, while losing only 0.88% of their original score for Pascal VOC 2007.