Understanding the Effective Receptive Field in Deep Convolutional Neural Networks

Abstract

We study characteristics of receptive fields of units in deep convolutional networks. The receptive field size is a crucial issue in many visual tasks, as the output must respond to large enough areas in the image to capture information about large objects. We introduce the notion of an effective receptive field, and show that it both has a Gaussian distribution and only occupies a fraction of the full theoretical receptive field. We analyze the effective receptive field in several architecture designs, and the effect of nonlinear activations, dropout, sub-sampling and skip connections on it. This leads to suggestions for ways to address its tendency to be too small.

Cited by in corpus (159)

• MLP-Mixer: An all-MLP Architecture for Vision
• SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
• Attentional Local Contrast Networks for Infrared Small Target Detection
• TransMorph: Transformer for unsupervised medical image registration
• Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification
• Gather-Excite: Exploiting Feature Context in Convolutional Neural Networks
• On the Compactness, Efficiency, and Representation of 3D Convolutional Networks: Brain Parcellation as a Pretext Task
• Dense Attention Fluid Network for Salient Object Detection in Optical Remote Sensing Images
• Systematic evaluation of CNN advances on the ImageNet
• A Review on Deep Learning Techniques for Video Prediction
• HED-UNet: Combined Segmentation and Edge Detection for Monitoring the Antarctic Coastline
• Learning Depth from Single Images with Deep Neural Network Embedding Focal Length
• A Deep Learning Interpretable Classifier for Diabetic Retinopathy Disease Grading
• ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias
• Compact Generalized Non-local Network
• Adversarial Networks for Spatial Context-Aware Spectral Image Reconstruction from RGB
• Receptive Field Block Net for Accurate and Fast Object Detection
• Explainable Deep One-Class Classification
• TransReID: Transformer-based Object Re-Identification
• Scale-Aware Trident Networks for Object Detection
• Knowledge Transfer between Buildings for Seismic Damage Diagnosis through Adversarial Learning
• DirectPose: Direct End-to-End Multi-Person Pose Estimation
• Deep Learning for Joint Channel Estimation and Feedback in Massive MIMO Systems
• Video Frame Interpolation via Adaptive Separable Convolution
• A newcomer's guide to deep learning for inverse design in nano-photonics
• Revisiting Light Field Rendering with Deep Anti-Aliasing Neural Network
• Neural Network-Based Automatic Liver Tumor Segmentation With Random Forest-Based Candidate Filtering
• SiamFC++: Towards Robust and Accurate Visual Tracking with Target Estimation Guidelines
• SNIPER: Efficient Multi-Scale Training
• Receptive Field Regularization Techniques for Audio Classification and Tagging with Deep Convolutional Neural Networks
• Adversarial Explanations for Understanding Image Classification Decisions and Improved Neural Network Robustness
• Efficient Neural Architecture Transformation Searchin Channel-Level for Object Detection
• Boosted $W/Z$ Tagging with Jet Charge and Deep Learning
• Compute and memory efficient universal sound source separation
• Global Aggregation then Local Distribution in Fully Convolutional Networks
• LatentGNN: Learning Efficient Non-local Relations for Visual Recognition
• LFFD: A Light and Fast Face Detector for Edge Devices
• Transporter Networks: Rearranging the Visual World for Robotic Manipulation
• Relation-Aware Global Attention for Person Re-identification
• PatchGuard: A Provably Robust Defense against Adversarial Patches via Small Receptive Fields and Masking
• Deformable Kernels: Adapting Effective Receptive Fields for Object Deformation
• S${}^3$FD: Single Shot Scale-invariant Face Detector
• Learning Better Features for Face Detection with Feature Fusion and Segmentation Supervision
• CyCNN: A Rotation Invariant CNN using Polar Mapping and Cylindrical Convolution Layers
• Mind the Pad -- CNNs can Develop Blind Spots
• A multi-task learning for cavitation detection and cavitation intensity recognition of valve acoustic signals
• Video Action Understanding
• Computation Reallocation for Object Detection
• Context-aware Synthesis for Video Frame Interpolation
• DDCNet: Deep Dilated Convolutional Neural Network for Dense Prediction
• Global Texture Enhancement for Fake Face Detection in the Wild
• Structured Attentions for Visual Question Answering
• Asymmetric 3D Context Fusion for Universal Lesion Detection
• Resolution learning in deep convolutional networks using scale-space theory
• Recent Advances in Deep Learning for Object Detection
• Feature Agglomeration Networks for Single Stage Face Detection
• Medical Image Segmentation Using Squeeze-and-Expansion Transformers
• FCHD: Fast and accurate head detection in crowded scenes
• Mechanisms of a Convolutional Neural Network for Learning Three-dimensional Unsteady Wake Flow
• Blurring the Line Between Structure and Learning to Optimize and Adapt Receptive Fields
• Global Aggregation then Local Distribution for Scene Parsing
• Improving Semantic Segmentation via Decoupled Body and Edge Supervision
• Bridge Damage Detection using a Single-Stage Detector and Field Inspection Images
• Efficient Visual Recognition with Deep Neural Networks: A Survey on Recent Advances and New Directions
• Spatially-Adaptive Filter Units for Compact and Efficient Deep Neural Networks
• Spatial Dual-Modality Graph Reasoning for Key Information Extraction
• Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentation
• Non-Intrusive Load Monitoring with Fully Convolutional Networks
• Insights into End-to-End Learning Scheme for Language Identification
• Spatially-Attentive Patch-Hierarchical Network for Adaptive Motion Deblurring
• Align Deep Features for Oriented Object Detection
• Funnel Activation for Visual Recognition
• CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching
• Efficient 3D Fully Convolutional Networks for Pulmonary Lobe Segmentation in CT Images
• Hard-Attention for Scalable Image Classification
• Hyperspectral Image Classification With Context-Aware Dynamic Graph Convolutional Network
• Look Before You Leap: Learning Landmark Features for One-Stage Visual Grounding
• NL-LinkNet: Toward Lighter but More Accurate Road Extraction with Non-Local Operations
• Compact Global Descriptor for Neural Networks
• Attribute Prototype Network for Zero-Shot Learning
• Scale-aware Automatic Augmentation for Object Detection
• Low-Complexity Models for Acoustic Scene Classification Based on Receptive Field Regularization and Frequency Damping
• flexgrid2vec: Learning Efficient Visual Representations Vectors
• ACNN: a Full Resolution DCNN for Medical Image Segmentation
• Geometric Style Transfer
• PARN: Position-Aware Relation Networks for Few-Shot Learning
• Progressively Guided Alternate Refinement Network for RGB-D Salient Object Detection
• FeatherNets: Convolutional Neural Networks as Light as Feather for Face Anti-spoofing
• Dynamic Context Correspondence Network for Semantic Alignment
• Positional Encoding as Spatial Inductive Bias in GANs
• Domain-invariant Stereo Matching Networks
• Discourse-level Relation Extraction via Graph Pooling
• Using neural networks to solve the 2D Poisson equation for electric field computation in plasma fluid simulations
• Patchnet: Interpretable Neural Networks for Image Classification
• Irregular Convolutional Neural Networks
• Integrating Large Circular Kernels into CNNs through Neural Architecture Search
• Learning Scene Gist with Convolutional Neural Networks to Improve Object Recognition
• A Feature Fusion-Net Using Deep Spatial Context Encoder and Nonstationary Joint Statistical Model for High Resolution SAR Image Classification
• Cheaper Pre-training Lunch: An Efficient Paradigm for Object Detection
• Multi-Resolution Fully Convolutional Neural Networks for Monaural Audio Source Separation
• Learning from Videos with Deep Convolutional LSTM Networks
• Boundary Aware U-Net for Glacier Segmentation
• End-to-end Interpretable Neural Motion Planner
• SegNBDT: Visual Decision Rules for Segmentation
• Discovering Dynamic Salient Regions for Spatio-Temporal Graph Neural Networks
• Deep Learning for Robust Motion Segmentation with Non-Static Cameras
• LIP: Local Importance-based Pooling
• Propose-and-Attend Single Shot Detector
• AFO-TAD: Anchor-free One-Stage Detector for Temporal Action Detection
• DeepFlux for Skeletons in the Wild
• Dynamic Filtering with Large Sampling Field for ConvNets
• Detecting Reflections by Combining Semantic and Instance Segmentation
• GridTracer: Automatic Mapping of Power Grids using Deep Learning and Overhead Imagery
• A Single-shot Object Detector with Feature Aggragation and Enhancement
• Representative Graph Neural Network
• Deep Convolutional Neural Network-based Bernoulli Heatmap for Head Pose Estimation
• Learning To Pay Attention To Mistakes
• Visual Concept Reasoning Networks
• 3D Neighborhood Convolution: Learning Depth-Aware Features for RGB-D and RGB Semantic Segmentation
• On Optimizing Deep Convolutional Neural Networks by Evolutionary Computing
• Sams-Net: A Sliced Attention-based Neural Network for Music Source Separation
• Segmentation of EM showers for neutrino experiments with deep graph neural networks
• Recurrent Attention Model with Log-Polar Mapping is Robust against Adversarial Attacks
• Inception Convolution with Efficient Dilation Search
• Subjective and Objective De-raining Quality Assessment Towards Authentic Rain Image
• Improving Skeleton-based Action Recognitionwith Robust Spatial and Temporal Features
• Deformable Kernel Convolutional Network for Video Extreme Super-Resolution
• Bayesian deep learning for mapping via auxiliary information: a new era for geostatistics?
• Self-supervised Learning with Fully Convolutional Networks
• A Spectral Nonlocal Block for Neural Networks
• PGT: A Progressive Method for Training Models on Long Videos
• Adaptive Context Network for Scene Parsing
• GCCN: Global Context Convolutional Network
• Signature-Graph Networks
• CASSOD-Net: Cascaded and Separable Structures of Dilated Convolution for Embedded Vision Systems and Applications
• Examining and Mitigating Kernel Saturation in Convolutional Neural Networks using Negative Images
• Real-Time Semantic Segmentation via Auto Depth, Downsampling Joint Decision and Feature Aggregation
• Boundary Guidance Hierarchical Network for Real-Time Tongue Segmentation
• Density-embedding layers: a general framework for adaptive receptive fields
• Efficient Transfer Learning via Joint Adaptation of Network Architecture and Weight
• Using LIP to Gloss Over Faces in Single-Stage Face Detection Networks
• MBS: Macroblock Scaling for CNN Model Reduction
• Multi-Scale Convolutions for Learning Context Aware Feature Representations
• Poly-NL: Linear Complexity Non-local Layers with Polynomials
• SAFE: Scale Aware Feature Encoder for Scene Text Recognition
• Approximate Random Dropout
• Learning Topology from Synthetic Data for Unsupervised Depth Completion
• Scaling up deep neural networks: a capacity allocation perspective
• A Coarse-to-Fine Instance Segmentation Network with Learning Boundary Representation
• Learnable Sampling 3D Convolution for Video Enhancement and Action Recognition
• A novel Deep Structure U-Net for Sea-Land Segmentation in Remote Sensing Images
• Log-Polar Space Convolution for Convolutional Neural Networks
• Encoder-Decoder based CNN and Fully Connected CRFs for Remote Sensed Image Segmentation
• ARMA Nets: Expanding Receptive Field for Dense Prediction
• Image denoising by Super Neurons: Why go deep?
• Enabling variable high spatial resolution retrieval from a long pulse BOTDA sensor
• LUAI Challenge 2021 on Learning to Understand Aerial Images
• MisMatch: Calibrated Segmentation via Consistency on Differential Morphological Feature Perturbations with Limited Labels
• Extra Proximal-Gradient Inspired Non-local Network