TreeNet: A lightweight One-Shot Aggregation Convolutional Network

by   Lu Rao, et al.

The architecture of deep convolutional networks (CNNs) has evolved for years, becoming more accurate and faster. However, it is still challenging to design reasonable network structures that aim at obtaining the best accuracy under a limited computational budget. In this paper, we propose a Tree block, named after its appearance, which extends the One-Shot Aggregation (OSA) module while being more lightweight and flexible. Specifically, the Tree block replaces each of the 3×3 Conv layers in OSA into a stack of shallow residual block (SRB) and 1×1 Conv layer. The 1×1 Conv layer is responsible for dimension increasing and the SRB is fed into the next step. By doing this, when aggregating the same number of subsequent feature maps, the Tree block has a deeper network structure while having less model complexity. In addition, residual connection and efficient channel attention(ECA) is added to the Tree block to further improve the performance of the network. Based on the Tree block, we build efficient backbone models calling TreeNets. TreeNet has a similar network architecture to ResNet, making it flexible to replace ResNet in various computer vision frameworks. We comprehensively evaluate TreeNet on common-used benchmarks, including ImageNet-1k for classification, MS COCO for object detection, and instance segmentation. Experimental results demonstrate that TreeNet is more efficient and performs favorably against the current state-of-the-art backbone methods.


page 1

page 2

page 3

page 4


EPSANet: An Efficient Pyramid Split Attention Block on Convolutional Neural Network

Recently, it has been demonstrated that the performance of a deep convol...

ResNetX: a more disordered and deeper network architecture

Designing efficient network structures has always been the core content ...

AutoBSS: An Efficient Algorithm for Block Stacking Style Search

Neural network architecture design mostly focuses on the new convolution...

Recurrence along Depth: Deep Convolutional Neural Networks with Recurrent Layer Aggregation

This paper introduces a concept of layer aggregation to describe how inf...

Res2Net: A New Multi-scale Backbone Architecture

Representing features at multiple scales is of great importance for nume...

Data-Driven Neuron Allocation for Scale Aggregation Networks

Successful visual recognition networks benefit from aggregating informat...

Concurrently Extrapolating and Interpolating Networks for Continuous Model Generation

Most deep image smoothing operators are always trained repetitively when...

Please sign up or login with your details

Forgot password? Click here to reset