Modeling local and global deformations in Deep Learning: Epitomic convolution, Multiple Instance Learning, and sliding window detection

By George Papandreou, Iasonas Kokkinos, Pierre-André Savalle. 2015

200 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures

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

In Model Architectures

Citations in the corpus

Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation (2020, methodology): “Some employ atrous convolutions [34,77,67,13], larger kernel [70], or image pyramids [98,85], either designed by hand or searched by algorithms [103,12,60].”
DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (2016, methodology): “A standard way to deal with this is to present to the DCNN rescaled versions of the same image and then aggregate the feature or score maps [6], [17], [18].”
Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation (2018, methodology): “In this section, we briefly introduce atrous convolution [30, 21, 64, 56, 8] and depthwise separable convolution [67, 71, 74, 12, 31].”
MobileNetV2: Inverted Residuals and Linear Bottlenecks (2018, methodology): “DeepLabv3 adopts atrous convolution [40, 41, 42], a powerful tool to explicitly control the resolution of computed feature maps, and builds five parallel heads including (a) Atrous Spatial Pyramid Pooling module (ASPP) [43] containing three 3 × 3 convolutions with different atrous rates, (b) 1 × 1 c”
Rethinking Atrous Convolution for Semantic Image Segmentation (2017, methodology): “Instead, we advocate the use of ‘atrous convolution’, originally developed for the efficient computation of the undecimated wavelet transform in the “algorithme à trous” scheme of [30] and used before in the DCNN context by [21, 64, 56].”
Searching for MobileNetV3 (2019, methodology): “We apply atrous convolution [18, 40, 33, 6] to the last block of MobileNetV3 to extract denser features, and further add a skip connection [30] from low-level features to capture”
Pushing the Boundaries of Boundary Detection using Deep Learning (2015, methodology): “…excellent results, and in particular the use of the Dense Conditional Random Field (DenseCRF) of Kr ¨ ahenb ¨ uhl & Koltun (2011) by Chen et al. (2015); Papandreou et al. (2015a); Zheng et al. (2015), has enhanced the discriminative power of FCNNs with local evidence gathered by the image…”
SPGNet: Semantic Prediction Guidance for Scene Parsing (2019, methodology): “Typical examples include DeepLab [7] which adopts several parallel atrous convolutions [22, 43] with different rates and PSPNet [73] which performs pooling operations at different grid scales.”
DeeperLab: Single-Shot Image Parser (2019, methodology): “Recently, PSPNet [36] proposes using spatial pyramid pooling [37, 38] at several grid scales (including imagelevel pooling [39]), and DeepLab [35, 40] proposes applying several parallel atrous convolutions [41, 42, 21, 43, 44] with different rates (called Atrous Spatial Pyramid Pooling, or ASPP).”
DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution (2020, methodology): “Atrous convolution [11,30,53] is an effective technique to enlarge the ﬁeld-of-view of ﬁlters at any convolutional layer.”
StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth Prediction (2018, methodology): “We use atrous convolutions in these blocks to sample from a larger context without increasing the network size [43].”
CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples (2016, background): “We assume that a network is fully convolutional [45] or that all fully connected layers are discarded.”
Beyond Planar Symmetry: Modeling Human Perception of Reflection and Rotation Symmetries in the Wild (2017, methodology): “Atrous convolution [5, 6, 23, 49] is useful to provide contextual information for each pixel.”
Scaling Wide Residual Networks for Panoptic Segmentation (2020, background): “The SAC operation essentially gathers features computed with different atrous rates [31, 65, 5].”