Caffe Deep Learning
Caffe deep learning refers to a powerful open source tool that enables to train and use deep learning networks that are able to work with big data. The framework has been developed by the Berkeley AI Research team. Please refer to our article on deep learning for more general pieces of information about this machine learning approach. This article provides a short overview of the key features using the Caffe deep learning framework. When you are interested in another set of deep learning tools have a look on our Deep Learning Framework list.
The Caffe framework is written in C++ with a Python interface and offers extensibility to other developers that want to contribute to it. The expressive architecture enables simple applications. In other words without hard-coding code different deep learning models and their optimization can be created by simple configuration steps. The framework can be used with GPUs and CPUs by changing a single configuration item. This tool offers one of the fastest Convolutional Neural Network implementation available as open source. There is a CaffeOnSpark in order to provide more parallelization features developed by Yahoo.
Since the creation of the tool as a PhD thesis many other contributors have added many features to the tool or implemented optimizations. For example there is a optimized Intel extension for CPU and support of multi-nodes in order to support Xeon processors like Xeon Phi. Also you can find an OpenCL Caffe extension for AMD or Intel devices. The official Web page of the tool can be found here. The source code can be found on GitHub here.It can be referenced as follows:
Jia, Yangqing, Evan Shelhamer, Jeff Donahue, Sergey Karayev, Jonathan Long, Ross Girshick, Sergio Guadarrama, and Trevor Darrell. "Caffe: Convolutional architecture for fast feature embedding." In Proceedings of the 22nd ACM international conference on Multimedia, pp. 675-678. ACM, 2014.
Caffe Deep Learning Details
The following video provides more details about this topic: