Sparsely Aggregated Convolutional Networks

By Ligeng Zhu, Ruizhi Deng, Michael Maire, Zhiwei Deng, Greg Mori, Ping Tan. 2018

66 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Hongpeng Liao
2. Chien-Yao Wang
3. Song Han
4. I-Hau Yeh
5. Han Cai
6. Ligeng Zhu
7. Quoc V. Le
8. Eldad Haber

Citations in the corpus

• ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware (2018, methodology): “2 EXPERIMENTS ON IMAGENET For ImageNet experiments, we focus on learning efficient CNN architectures (Iandola et al., 2016; Howard et al., 2017; Sandler et al., 2018; Zhu et al., 2018) that have not only high accuracy but also low latency on specific hardware platforms.”
• Deep Residual-Dense Lattice Network for Speech Enhancement (2020, background): “While RDL blocks do not allow for total feature re-usage, densely aggregating only a subset of previous outputs has been shown to be beneficial (Zhu et al. 2018).”
• COVID-19 Detection Through Transfer Learning Using Multimodal Imaging Data (2020, methodology): “Some recent substantial models have been developed based on the base models, such as Res2Net [83] and Wide ResNet [84] using the ResNet model; while Log Dense Net [85] and Sparse Net [86] using the DenseNet model.”
• Multi-level Residual Networks from Dynamical Systems View (2017, background): “SparseNets (Zhu et al., 2018) is a simple feature aggregation structure with shortcut paths bypassing exponentially growing number of layers.”
• An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection (2019, unknown): “[26],”
• HarDNet: A Low Memory Traffic Network (2019, methodology): “Therefore, we implemented a FC-SparseNet-ref100 for comparison and trained it in the same environment for five runs, and then we picked the best result.”
• CSPNet: A New Backbone that can Enhance Learning Capability of CNN (2019, methodology): “SparseNet [46] adjusts dense connection to exponentially spaced connection can effectively improve parameter utilization and thus result in better outcomes.”
• Designing Network Design Strategies Through Gradient Path Analysis (2022, methodology): “It can be seen from Figure 8 that the concatenation connection-based architectures, such as DenseNet and SparseNet [39], belong to the network that must be specially handled.”