An Investigation on Performance of Attention Deep Neural Networks in Rapid Object Recognition
- 39 Downloads
Inspite of the huge success of Deep Neural Networks in object recognition, there are still situations in which they cannot reach human performance. In this work, the performance of an attention Deep Neural Network which is cued by important pixel of objects is addressed. First, the effect of color on accuracy of classification is evaluated. Then the network performance is compared with humans by using a set of images from different levels of revelation of important pixels. The results indicate that color information enhances the recognition of objects and there is a correspondence in accuracy of classification as well as correlation in decisions between human and attention networks at middle and low levels of important pixel revelation respectively.
KeywordsObject recognition Attention Deep neural networks
This research has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement No 665919. Thanks to Thomas Serre and Drew Linsley.
- 5.Eberhardt, S., Cader, J.G., Serre, T.: How deep is the feature analysis underlying rapid visual categorization? In: Advances in Neural Information Processing Systems, pp. 1100–1108 (2016)Google Scholar
- 7.Geirhos, R., Janssen, D.H., Schütt, H.H., Rauber, J., Bethge, M., Wichmann, F.A.: Comparing deep neural networks against humans: object recognition when the signal gets weaker. arXiv preprint arXiv:1706.06969 (2017)
- 8.Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7132–7141 (2018)Google Scholar
- 10.Linsley, D., Scheibler, D., Eberhardt, S., Serre, T.: Global-and-local attention networks for visual recognition. arXiv preprint arXiv:1805.08819 (2018)