Caffe

Summary

Caffe is a fast, expressive, and extensible deep learning framework developed by Berkeley AI Research and community contributors.

Abstract

Caffe is a deep learning framework designed for speed, modularity, and expressiveness in application and innovation. It was created by Yangqing Jia during his PhD at UC Berkeley and is released under the BSD 2-Clause license. Caffe's extensible code has been contributed to by over 1,000 developers, and it powers numerous academic, startup, and industrial applications. The framework offers documentation, tutorials, and examples for learning and development. It supports both CPU and GPU processing, and can handle large-scale data for research and industry deployment.

Bullet Points

• •Deep learning framework developed by Berkeley AI Research and community contributors
• •Created by Yangqing Jia during his PhD at UC Berkeley and released under BSD 2-Clause license
• •Expressive architecture encourages application and innovation with models and optimization defined by configuration
• •Extensible code fosters active development with over 1,000 developers contributing
• •Processes over 60M images per day with a single NVIDIA K40 GPU
• •Supports both CPU and GPU processing for research and industry deployment
• •Provides documentation, tutorials, and examples for learning and development
• •Powers numerous academic, startup, and industrial applications
• •Offers Model Zoo with a standard distribution format for Caffe models and trained models
• •Supports fine-tuning for style recognition, multilabel classification, and off-the-shelf SGD for classification
• •Provides API documentation and benchmarking for different networks and GPUs
• •Community support available through caffe-users group and Github