Deep Layer Aggregation

Abstract

Visual recognition requires rich representations that span levels from low to high, scales from small to large, and resolutions from fine to coarse. Even with the depth of features in a convolutional network, a layer in isolation is not enough: compounding and aggregating these representations improves inference of what and where. Architectural efforts are exploring many dimensions for network backbones, designing deeper or wider architectures, but how to best aggregate layers and blocks across a network deserves further attention. Although skip connections have been incorporated to combine layers, these connections have been "shallow" themselves, and only fuse by simple, one-step operations. We augment standard architectures with deeper aggregation to better fuse information across layers. Our deep layer aggregation structures iteratively and hierarchically merge the feature hierarchy to make networks with better accuracy and fewer parameters. Experiments across architectures and tasks show that deep layer aggregation improves recognition and resolution compared to existing branching and merging schemes. The code is at https://github.com/ucbdrive/dla.

Published at the Conference on Computer Vision and Pattern Recognition (CVPR) 2018

References in corpus (2)

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• How transferable are features in deep neural networks?

Cited by in corpus (14)

• AutoAugment: Learning Augmentation Policies from Data
• FishNet: A Versatile Backbone for Image, Region, and Pixel Level Prediction
• Saliency for Fine-grained Object Recognition in Domains with Scarce Training Data
• DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation
• IDK Cascades: Fast Deep Learning by Learning not to Overthink
• SMOKE: Single-Stage Monocular 3D Object Detection via Keypoint Estimation
• Towards Ghost-free Shadow Removal via Dual Hierarchical Aggregation Network and Shadow Matting GAN
• Disentangling Propagation and Generation for Video Prediction
• Glance and Gaze: Inferring Action-aware Points for One-Stage Human-Object Interaction Detection
• Sparsely Aggregated Convolutional Networks
• Multi-scale Convolution Aggregation and Stochastic Feature Reuse for DenseNets
• Accurate Nuclear Segmentation with Center Vector Encoding
• Dense Fusion Classmate Network for Land Cover Classification
• Learning to see across Domains and Modalities