Designing Deep Networks for Scene Recognition
This work addresses the problem of inefficient network design for scene recognition tasks, offering a domain-specific improvement that is incremental in nature.
The paper tackles the problem of designing deep networks for scene recognition by showing that widely accepted network design principles can lead to dramatic performance differences when data characteristics change, and it proposes a data-oriented methodology with a Deep-Narrow Network and Dilated Pooling module that improves scene recognition performance using less than half the computational resources compared to ResNets.
Most deep learning backbones are evaluated on ImageNet. Using scenery images as an example, we conducted extensive experiments to demonstrate the widely accepted principles in network design may result in dramatic performance differences when the data is altered. Exploratory experiments are engaged to explain the underlining cause of the differences. Based on our observation, this paper presents a novel network design methodology: data-oriented network design. In other words, instead of designing universal backbones, the scheming of the networks should treat the characteristics of data as a crucial component. We further proposed a Deep-Narrow Network and Dilated Pooling module, which improved the scene recognition performance using less than half of the computational resources compared to the benchmark network architecture ResNets. The source code is publicly available on https://github.com/ZN-Qiao/Deep-Narrow-Network.