Caffe Introduction
Caffe website CS231n Caffe Tutorial
Structure of Caffe
- **Blob**: Stores data and derivatives
- **Layer**: Transforms bottom blobs to top blobs
- **Net**: Many layers; computes gradients via forward backward
- **Solver**: Uses gradients to update weights
Steps to run Caffe
- Convert the data to Caffe-format lmdb, leveldb, hdf5 / .mat, list of images, etc.
- Define the Net
- Configure the Solver
- caffe train -solver solver.prototxt -gpu 0
How to do fine-tuning
- Download the train and val data
- Transfer them as the input style (jpeg to txt to HDF5)
Register on VOC and Download the test data and transfer
Make a network structure
Make a solver
Run ./build/tools/caffe train -solver models/VGG_ILSVRC_16_layers/solver.prototxt -weights models/VGG_ILSVRC_16_layers/VGG_ILSVRC_16_layers.caffemodel -gpu 0
Fine-tuning CaffeNet for Style Recognition on “Flickr Style” Data