Home

What is Caffe?

Caffe is an open-source implementation of the recent convolutional algorithms that perform particularly well in large-scale image recognition tasks, such as the ImageNet Challenges. The purpose of Caffe is to provide a reference implementation for such algorithms, contribute a framework for developing and deploying deep learning architectures for vision, and enable wider adoption and analysis in the research community.

Caffe was created by Yangqing Jia at UC Berkeley as a replacement of decaf. It has been adopted by several Berkeley vision group members and is under active development by the Berkeley Vision and Learning Center.