Dual Path Networks

By Yunpeng Chen, Jianan Li, Huaxin Xiao, Xiaojie Jin, Shuicheng Yan, Jiashi Feng. 2017

903 citations recorded by the source corpus.

In this work, we present a simple, highly efficient and modularized Dual Path Network (DPN) for image classification which presents a new topology of connection paths internally. By revealing the equivalence of the state-of-the-art Residual Network (ResNet) and Densely Convolutional Network (DenseNet) within the HORNN framework, we find that ResNet enables feature re-usage while DenseNet enables new features exploration which are both important for learning good representations. To enjoy the benefits from both path topologies, our proposed Dual Path Network shares common features while maintaining the flexibility to explore new features through dual path architectures. Extensive experiments on three benchmark datasets, ImagNet-1k, Places365 and PASCAL VOC, clearly demonstrate superior performance of the proposed DPN over state-of-the-arts. In particular, on the ImagNet-1k dataset, a shallow DPN surpasses the best ResNeXt-101(64x4d) with 26% smaller model size, 25% less computational cost and 8% lower memory consumption, and a deeper DPN (DPN-131) further pushes the state-of-the-art single model performance with about 2 times faster training speed. Experiments on the Places365 large-scale scene dataset, PASCAL VOC detection dataset, and PASCAL VOC segmentation dataset also demonstrate its consistently better performance than DenseNet, ResNet and the latest ResNeXt model over various applications.

Subfields: Civil and Structural Engineering, Computer Vision and Pattern Recognition, Urban Studies

How researchers see this paper

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

In Civil and Structural Engineering

In Computer Vision and Pattern Recognition

In Urban Studies

Citations in the corpus

Measuring Robustness to Natural Distribution Shifts in Image Classification (2020, neutral): “…includes models trained on (i) Facebook’s collection of 1 billion Instagram images [56, 104], (ii) the YFCC 100 million dataset [104], (iii) Google’s JFT 300 million dataset [82, 102], (iv) a subset of OpenImages [98], or (v) a subset of the full ImageNet dataset of 21,841 classes [11, 49, 99].”
Natural Adversarial Examples (2019, background): “Figure 10: Increasing the capacity of ResNets, DualPathNetworks [10], and ResNeXts improves accuracy on IMAGENET-A.”
Do ImageNet Classifiers Generalize to ImageNet? (2019, background): “dpn107 [4] https://github.”
VOLO: Vision Outlooker for Visual Recognition (2021, background): “DPNs [7] leverage both residual and dense connections for designing stronger building blocks.”
Squeeze-and-Excitation Networks (2017, background): “Reformulations of the connections between network layers [5, 14] have been shown to further improve the learning and representational properties of deep networks.”
Green AI (2019, neutral): “(4a): Different models: AlexNet [20], ResNet [14], ResNext [46], DPN107 [6], SENet154 [17].”
Hybrid Task Cascade for Instance Segmentation (2019, positive): “We utilize an emsemble of ﬁve networks: SENet-154 [19], ResNeXt-101 [43] 64x4d, ResNeXt-101 32x8d, DPN-107 [9], FishNet [40].”
Progressive Neural Architecture Search (2017, neutral): “ication results in the Mobile setting. Model Image Size Params Mult-Adds Top-1 Top-5 ResNeXt-101 (64x4d) [36] 320 320 83.6M 31.5B 80.9 95.6 PolyNet [38] 331 331 92M 34.7B 81.3 95.8 Dual-Path-Net-131 [6] 320 320 79.5M 32.0B 81.5 95.8 Squeeze-Excite-Net [15] 320 320 145.8M 42.3B 82.7 96.2 NASNet-A (N=”
Regularized Evolution for Image Classifier Architecture Search (2018, background): “Until recently, most state-of-the-art (SOTA) convolutional architectures have been manually designed by human experts (Krizhevsky et al., 2012; Ciregan et al., 2012; Lin et al., 2014; Sermanet et al., 2014; He et al., 2014; Chatfield et al., 2014; Simonyan & Zisserman, 2015; Karpathy et al., 2014; S”
Path Aggregation Network for Instance Segmentation (2018, background): “Increasing the flexibility and diversity of information paths by creating parallel paths following the splittransform-merge strategy [61, 6] is also beneficial.”
Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions (2021, methodology): “The effectiveness of a multi-path convolutional block was further validated in Inception series [60, 58], ResNeXt [73], DPN [10], MixNet [65] and SKNet [36].”
A Comprehensive Survey of Neural Architecture Search (2020, neutral): “1/Top 5 Acc(%) Params (Millions) Image Size (squarred) GPU Days Human Mobilenets [6] CoRR17 70.6/89.5 4.2 224 - ResNeXt [138] CVPR17 80.9/95.6 83.6 320 - Polynet [139] CVPR17 81.3/95.8 92.0 331 - DPN [140] NIPS17 81.5/95.8 79.5 320 - Shufflenet [137] CVPR18 70.9/89.8 5.0 224 - RL NASNet [31] CVPR18 ”
Peephole: Predicting Network Performance Before Training (2017, neutral): “Many representative architectures, such as AlexNet [13], VGGNet [21], GoogLeNet [23], ResNet [7], DenseNet [9], and DPN [3] are designed manually, based on intuitions and experience.”
Learned Video Compression (2018, methodology): “Specifically, we combine the multiscale rendition [19] of DenseNet [20] with dual path [9].”
CornerNet: Detecting Objects as Paired Keypoints (2018, neutral): “2 - - - - - D-RFCN + SNIP (Singh and Davis, 2017) DPN-98 (Chen et al., 2017) 45.”
Understanding Batch Normalization (2018, neutral): “This is expected since the depth of the network "ahead" of lower layers are deeper for the early layers, which would cause them to have worse conditioning according to previous observations in RMT.”