Backbone Network For Vision

Architecture design is one of the most important research topics in deep learning. Since the launch of AlexNet, convolutional neural networks have become the dominant machine learning approach for computer vision. Milestone works, such as ResNet, DenseNet, have greatly contributed to the development of the field. Recently, vision transformers, such as ViT, exhibit promising capability for vision tasks. In LEAP Lab, we focus on designing efficient and effective visual architectures.

Highlighted Work:

Densely Connected Convolutional Networks

Gao Huang, Zhuang Liu, Laurens van der Maaten, Kilian Q. Weinberger

CVPR 2017 (Best Paper Award)   ·   30 Jan 2018   ·   arxiv:1608.06993

Paper

Code