#ai #machinelearning #attention

Convolutional Neural Networks have dominated image processing for the last decade, but transformers are quickly replacing traditional models. This paper proposes a fully attentional model for images by combining learned Positional Embeddings with Axial Attention. This new model can compete with CNNs on image classification and achieve state-of-the-art in various image segmentation tasks.

OUTLINE:
0:00 - Intro & Overview
4:10 - This Paper's Contributions
6:20 - From Convolution to Self-Attention for Images
16:30 - Learned Positional Embeddings
24:20 - Propagating Positional Embeddings through Layers
27:00 - Traditional vs Position-Augmented Attention
31:10 - Axial Attention
44:25 - Replacing Convolutions in ResNet
46:10 - Experimental Results & Examples

Paper: https://arxiv.org/abs/2003.07853
Code: https://github.com/csrhddlam/axial-de...

My Video on BigBird:    • Big Bird: Transformers for Longer Seq...  
My Video on ResNet:    • [Classic] Deep Residual Learning for ...  
My Video on Attention:    • Attention Is All You Need  

Abstract:
Convolution exploits locality for efficiency at a cost of missing long range context. Self-attention has been adopted to augment CNNs with non-local interactions. Recent works prove it possible to stack self-attention layers to obtain a fully attentional network by restricting the attention to a local region. In this paper, we attempt to remove this constraint by factorizing 2D self-attention into two 1D self-attentions. This reduces computation complexity and allows performing attention within a larger or even global region. In companion, we also propose a position-sensitive self-attention design. Combining both yields our position-sensitive axial-attention layer, a novel building block that one could stack to form axial-attention models for image classification and dense prediction. We demonstrate the effectiveness of our model on four large-scale datasets. In particular, our model outperforms all existing stand-alone self-attention models on ImageNet. Our Axial-DeepLab improves 2.8% PQ over bottom-up state-of-the-art on COCO test-dev. This previous state-of-the-art is attained by our small variant that is 3.8x parameter-efficient and 27x computation-efficient. Axial-DeepLab also achieves state-of-the-art results on Mapillary Vistas and Cityscapes.

Authors: Huiyu Wang, Yukun Zhu, Bradley Green, Hartwig Adam, Alan Yuille, Liang-Chieh Chen

Links:
YouTube:    / yannickilcher  
Twitter:   / ykilcher  
Discord:   / discord  
BitChute: https://www.bitchute.com/channel/yann...
Minds: https://www.minds.com/ykilcher
Parler: https://parler.com/profile/YannicKilcher
LinkedIn:   / yannic-kilcher-488534136  

If you want to support me, the best thing to do is to share out the content :)

If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
SubscribeStar: https://www.subscribestar.com/yannick...
Patreon:   / yannickilcher  
Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq
Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2
Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m
Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n