Caffe
In plain wordsAn early free deep-learning toolkit from the University of California, Berkeley, popular for image recognition in the mid-2010s, in which networks were described in simple text files rather than code.
An open-source deep-learning framework written in C++ with CUDA GPU support and Python and MATLAB interfaces, developed at the Berkeley Vision and Learning Center. Networks were defined declaratively as stacks of layers in plain-text configuration files ('prototxt' files in Protocol Buffers text format) and trained by a separately configured solver, so standard convolutional networksconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects. could be trained without programming. Fast GPU convolution and a Model Zoo of shared trained weights made Caffe a standard tool for computer vision in 2014–2016, used to train and distribute versions of models such as AlexNetAlexNet, prop. n.: The neural network that won a major image-recognition contest by a huge margin in 2012, convincing researchers and companies that deep learning trained on graphics chips was the way forward., VGG and GoogLeNet. Its rigid layer abstraction made recurrent networks and novel research ideas awkward, and it was overtaken by TensorFlowTensorFlow, prop. n.: Google's free software toolkit for building, training and running machine-learning models; it led the field after 2015 and still runs many models inside phones, browsers and company systems. and PyTorchPyTorch, prop. n.: The most widely used free software toolkit for building and training neural networks, especially in research; programmers write ordinary Python and it handles the calculus and the graphics chips.. Facebook's successor, Caffe2 (2017), targeted production and mobile deployment and was merged into PyTorch in 2018.
The old detection model is a Caffe prototxt, so convert it before trying to fine-tune it in PyTorch.
Origin
Created by Yangqing Jia during his PhD at the University of California, Berkeley, in 2013 and described with colleagues in a 2014 paper; the name stands for Convolutional Architecture for Fast Feature Embedding. Jia later led the development of Caffe2 at Facebook.
Related
- PyTorchPyTorch, prop. n.: The most widely used free software toolkit for building and training neural networks, especially in research; programmers write ordinary Python and it handles the calculus and the graphics chips.
- TensorFlowTensorFlow, prop. n.: Google's free software toolkit for building, training and running machine-learning models; it led the field after 2015 and still runs many models inside phones, browsers and company systems.
- TheanoTheano, prop. n.: One of the first popular free software tools for deep learning, built at a Montreal university lab; development ended in 2017, but it shaped the toolkits that replaced it.
- convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects.
- AlexNetAlexNet, prop. n.: The neural network that won a major image-recognition contest by a huge margin in 2012, convincing researchers and companies that deep learning trained on graphics chips was the way forward.
- ImageNetImageNet, prop. n.: A huge collection of more than 14 million labeled photos, sorted into object categories, whose yearly recognition contest from 2010 to 2017 sparked the deep learning boom.
Sources
- Jia, Y. et al. (2014). Caffe: Convolutional Architecture for Fast Feature Embedding. Proceedings of the 22nd ACM International Conference on Multimedia. arXiv:1408.5093.
Mentioned in
TheanoTheano, prop. n.: One of the first popular free software tools for deep learning, built at a Montreal university lab; development ended in 2017, but it shaped the toolkits that replaced it.