Wider or Deeper: Revisiting the ResNet Model for Visual Recognition

By Zifeng Wu, Chunhua Shen, Anton van den Hengel. 2016

1,731 citations recorded by the source corpus.

Abstract The community has been going deeper and deeper in designing one cutting edge network after another, yet some works are there suggesting that we may have gone too far in this dimension. Some researchers unravelled a residual network into an exponentially wider one, and assorted the success of residual networks to fusing a large amount of relatively shallow models. Since some of their early claims are still not settled, we in this paper dig more on this topic, i.e., the unravelled view of residual networks. Based on that, we try to find a good compromise between the depth and width. Afterwards, we walk through a typical pipeline of developing a deep-learning-based algorithm. We start from a group of relatively shallow networks, which perform as well or even better than the current (much deeper) state-of-the-art models on the ImageNet classification dataset. Then, we initialize fully convolutional networks (FCNs) using our pre-trained models, and tune them for semantic image segmentation. Results show that the proposed networks, as pre-trained features, can boost existing methods a lot. Even without exhausting the sophistical techniques to improve the classic FCN model, we achieve comparable results with the best performers on four widely-used datasets, i.e., Cityscapes, PASCAL VOC, ADE20k and PASCAL-Context. The code and pre-trained models are released for public access 1 .

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

A Constructive Prediction of the Generalization Error Across Scales (2019, methodology): “For all other datasets we use WRN-44-16 (Wu et al., 2016) of varying widths, modified from the implementation of Hoffer et al.”
The Reversible Residual Network: Backpropagation Without Storing Activations (2017, background): “ResNets currently represent the state-of-the-art in object recognition [11], semantic segmentation [33] and image generation [30].”
Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation (2018, background): “ Visualization results on val set. The last row shows a failure mode. Backbone Decoder ASPP Image-Level mIOU X-65 X X 77.33 X-65 X X X 78.79 X-65 X X 79.14 X-71 X X 79.55 Method Coarse mIOU ResNet-38 [83] X 80.6 PSPNet [24] X 81.2 Mapillary [86] X 82.0 DeepLabv3 X 81.3 DeepLabv3+ X 82.1 (a) val set ”
Rethinking Atrous Convolution for Semantic Image Segmentation (2017, neutral): “, networks like [27, 74]) due to limited GPU memory and thus it is usually applied during the inference stage [14].”
MaX-DeepLab: End-to-End Panoptic Segmentation with Mask Transformers (2020, positive): “MaX-DeepLab-L stacks an L = 2 decoder on top of Wide-ResNet-41 [102, 95, 11].”
Incremental Learning in Semantic Segmentation from Image Labels (2021, methodology): “We employ Deeplab V3 [13] architecture for all the experiments, with a ResNet-101 [27] backbone with output stride equal to 16 for Pascal VOC and a Wide-ResNet38 [65] with output stride 8 for COCO, both pre-trained on ImageNet.”
Deep High-Resolution Representation Learning for Visual Recognition (2019, neutral): “Dilated-ResNet-101 76:6 56:2 89:6 77:8 LKM [89] ResNet-152 76:9 - - - DUC-HDC [118] - 77 :6 53 6 90 1 75 2 SAC [139] Dilated-ResNet-101 78:1 - - - DepthSeg [56] Dilated-ResNet-101 78:2 - - - ResNet38 [123] WResNet-38 78:4 59:1 90:9 78:1 BiSeNet [133] ResNet-101 78:9 - - - DFN [134] ResNet-101 79:3 -”
Data-Driven Sparse Structure Selection for Deep Neural Networks (2017, background): “Their explanation is also evidenced by subsequent work in network designs [37, 38].”
Semantic Instance Segmentation with a Discriminative Loss Function (2017, methodology): “The effect of the semantic segmentation is the largest: comparing the ﬁrst row (our method) to the third row, we see a large performance increase when replacing the ResNet-38 [40] semantic segmentation masks with the ground truth masks.”
Image Segmentation Using Deep Learning: A Survey (2020, neutral): “Wide ResNet [121], Exfuse [122] (enhancing low-level and high-level features fusion), Feedforward-Net [123],”