The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation

By Simon Jégou, Michal Drozdzal, David Vázquez, Adriana Romero, Yoshua Bengio. 2016

1,668 citations recorded by the source corpus.

State-of-the-art approaches for semantic image segmentation are built on Convolutional Neural Networks (CNNs). The typical segmentation architecture is composed of (a) a downsampling path responsible for extracting coarse semantic features, followed by (b) an upsampling path trained to recover the input image resolution at the output of the model and, optionally, (c) a post-processing module (e.g. Conditional Random Fields) to refine the model predictions.,,,,,, Recently, a new CNN architecture, Densely Connected Convolutional Networks (DenseNets), has shown excellent results on image classification tasks. The idea of DenseNets is based on the observation that if each layer is directly connected to every other layer in a feed-forward fashion then the network will be more accurate and easier to train.,,,,,, In this paper, we extend DenseNets to deal with the problem of semantic segmentation. We achieve state-of-the-art results on urban scene benchmark datasets such as CamVid and Gatech, without any further post-processing module nor pretraining. Moreover, due to smart construction of the model, our approach has much less parameters than currently published best entries for these datasets.

Subfields: Computer Vision and Pattern Recognition, Ocean Engineering

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

In Ocean Engineering

Citations in the corpus

Graph Attention Networks (2017, methodology): “Convolutional Neural Networks (CNNs) have been successfully applied to tackle problems such as image classiﬁcation (He et al., 2016), semantic segmentation (J´egou et al., 2017) or machine translation (Gehring et al., 2016), where the underlying data representation has a grid-like structure.”
Elucidating image-to-set prediction: An analysis of models, losses and datasets (2019, methodology): “Major advances in image understanding tasks [57, 18, 50, 49, 37, 23, 17] have been enabled by the introduction of large scale datasets such as ImageNet [53], MSCOCO [30] or Cityscapes [10].”
NAS-Unet: Neural Architecture Search for Medical Image Segmentation (2019, methodology): “The baseline methods are U-Net [8], FC-Densenet [27].”
MONet: Unsupervised Scene Decomposition and Representation (2019, background): “Neural architectures such as the U-Net [Ronneberger et al., 2015], Fully Convolutional DenseNet [Jégou et al., 2017], and DeepLabv3 [Chen et al., 2018]) have successively advanced the state of the art on this domain, furnishing strong inductive biases for image segmentation.”
What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? (2017, methodology): “Our model based on DenseNet [20] can process a 640 × 480 resolution image in 150 ms on a NVIDIA Titan X GPU.”
Concrete Dropout (2017, methodology): “We demonstrate Concrete dropout’s efﬁcacy by applying it to the DenseNet model [17] for semantic segmentation (example input, output, and uncertainty map is given in Figure 7).”
Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding (2015, methodology): “Moreover, we show that this technique is broadly applicable across a number of state of the art architectures and achieves a 2-3% improvement in segmenation accuracy when applied to SegNet [2], FCN [20], Dilation Network [30] and DenseNet [15].”
Accurate Uncertainties for Deep Learning Using Calibrated Regression (2018, methodology): “We use the DenseNet architecture of J´egou et al. (2017) for the depth regression task.”
Spherical CNNs on Unstructured Grids (2019, background): “…Segmentation Image semantic segmentation is a classic problem in computer vision, and there has been an impressive body of literature studying semantic segmentation of planar images (Ronneberger et al., 2015; Badrinarayanan et al., 2015; Long et al., 2015; J´egou et al., 2017; Wang et al., 2018a).”
Guide Me: Interacting with Deep Networks (2018, background): “Semantic segmentation In this paper, we focus on the application of semantic image segmentation, which is widely studied in the computer vision literature and significantly boosted by the success of deep learning methods [6, 12, 31, 39, 41, 51].”
CPFNet: Context Pyramid Fusion Network for Medical Image Segmentation (2020, methodology): “state-of-the-art algorithms, including PSPNet [23], DFN [19] TriamisuNet [41], cGAN [42], Attention U-Net [21], FastFCN [17], GCN [20], CE-Net [26] and our Baseline.”
Density-Aware Single Image De-raining Using a Multi-stream Dense Network (2018, background): “The transition layer can function as up-sample transition, downsample transition or no-sampling transition [14].”
Stacked Deconvolutional Network for Semantic Segmentation (2017, methodology): “Following [17], [29], [41], Mean IoU and Global Avg are employed to evaluate our method on this dataset.”
Intra-class Feature Variation Distillation for Semantic Segmentation (2020, neutral): “Params (M) Some related semantic segmentation methods FCN [25] 57.0 134.5 ENet [28] 51.3 0.3580 ESPNet [26] 57.8 0.3635 FC-DenseNet56 [22] 58.9 1.550 SegNet [4] 55.”
Pyramid Attention Network for Semantic Segmentation (2018, background): “Some kind of U-shape networks, such as SegNet[1], Refinenet[13], Tiramisu proposed in [10] and Large Kernel Matters[22] perform complicate decoder module which use low-level information to help high-level features recover images detail.”
IterNet: Retinal Image Segmentation Utilizing Structural Redundancy in Vessel Networks (2019, background): “Image segmentation: Currently, most state-of-the-art models [8, 9, 10] for semantic segmentation stem from a fully-convolutional design, which is first introduced by the fully convolutional network (FCN) [11].”
Survey on semantic segmentation using deep learning techniques (2019, background): “Jegou et al. [51] proposed a Fully Convolutional DenseNet FC-DenseNet, the extension of [71] by adding an upsampling path and skipping connections to recover the full input resolution.”
SEN12MS - A Curated Dataset of Georeferenced Multi-Spectral Sentinel-1/2 Imagery for Deep Learning and Data Fusion (2019, methodology): “The second network we used was the fully convolutional DenseNet for semantic segmentation (Jégou et al., 2017), aiming at assigning a class label to every pixel of the input image.”
Exascale Deep Learning for Climate Analytics (2018, methodology): “The first is a modification of the Tiramisu network [7].”
DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation (2019, positive): “The idea of dense connections has been recently proposed for image classification in [13] and extended to semantic segmentation in [14] [28].”