FBNet

Introduced by Wu et al. in FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search

FBNet is a type of convolutional neural architectures discovered through DNAS neural architecture search. It utilises a basic type of image model block inspired by MobileNetv2 that utilises depthwise convolutions and an inverted residual structure (see components).

Source: FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search

Read Paper See Code

Papers

Paper	Code	Results	Date	Stars

Tasks

Task	Papers	Share
Image Classification	3	18.75%
Semantic Segmentation	2	12.50%
Meta-Learning	1	6.25%
Point Cloud Completion	1	6.25%
Autonomous Driving	1	6.25%
Image Segmentation	1	6.25%
Scene Segmentation	1	6.25%
Quantization	1	6.25%
Depth Estimation	1	6.25%

Usage Over Time

This feature is experimental; we are continuously improving our matching algorithm.

Components

Component	Type	Add Remove
1x1 Convolution	Convolutions
Convolution	Convolutions
Dense Connections	Feedforward Networks
Dropout	Regularization
FBNet Block	Image Model Blocks
Global Average Pooling	Pooling Operations
Kaiming Initialization	Initialization
Softmax	Output Functions

Categories

Add Remove

Convolutional Neural Networks

Light-weight neural networks