The Google Research Brain Team Introduces a New Architecture for Computer Vision: MLP-Mixer
The Google Research Brain team has introduced a new architecture for computer vision, MLP-Mixer. Recently, attention-based networks, such as the Vision Transformer, have gained popularity. In this post, the Google Research Brain team presents MLP-Mixer, an architecture based solely on multilayer perceptrons (MLPs).
The Google Research Brain Team Introduces a New Architecture for Computer Vision: MLP-Mixer
Recently, attention-based networks, such as the Vision Transformer, have gained popularity. In this post, the Google Research Brain team presents the MLP-Mixer, an architecture based solely on multilayer perceptrons (MLPs). The MLP-Mixer comprises two types of layers: one with MLPs applied independently to image patches (i.e., "mixing" features by location) and another with MLPs across patches (i.e., "mixing" spatial information). When trained on large datasets, the MLP-Mixer achieved results comparable to the latest models. The team of scientists hopes that these results will inspire further research beyond well-established CNNs and transformers.
Source:
Originally published on Facebook — link to post
Původní zdroj: facebook