Non-Local Block

Introduced by Wang et al. in Non-local Neural Networks

A Non-Local Block is an image block module used in neural networks that wraps a non-local operation. We can define a non-local block as:

$$ \mathbb{z}_{i} = W_{z}\mathbb{y_{i}} + \mathbb{x}_{i} $$

where $y_{i}$ is the output from the non-local operation and $+ \mathbb{x}_{i}$ is a residual connection.

Source: Non-local Neural Networks

Read Paper See Code

Paper	Code	Results	Date	Stars

Task	Papers	Share
Image Generation	36	13.19%
Conditional Image Generation	15	5.49%
Object Detection	11	4.03%
Instance Segmentation	10	3.66%
Semantic Segmentation	10	3.66%
Super-Resolution	7	2.56%
Multi-agent Reinforcement Learning	7	2.56%
Reinforcement Learning (RL)	6	2.20%
Decision Making	5	1.83%

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

Component	Type	Add Remove
1x1 Convolution	Convolutions
Concatenation Affinity	Affinity Functions	(optional)
Embedded Dot Product Affinity	Affinity Functions	(optional)
Embedded Gaussian Affinity	Affinity Functions	(optional)
Gaussian Affinity	Affinity Functions	(optional)
Non-Local Operation	Image Feature Extractors
Residual Connection	Skip Connections